Life History Strategy and Ecosystem Impact of a Dominant Small Mammal Herbivore in a Mountain Steppe

Life history strategy and ecosystem impact of a dominant small mammal herbivore in a mountain steppe

Karin Nadrowski

12th April 2006

Contents

Acknowledgements ...... vii Abstract ...... viii Zusammenfassung ...... x

1 Introduction 1 1.1 Mountain steppe ...... 1 1.2 Herbivores ...... 2 1.3 Life history strategies ...... 3 1.4 From keystone species to small mammal pest: Ecosystem impacts . . . 4 1.5 Situation in Mongolia and project background ...... 7 1.6 Aim of this study ...... 8

2 Basic Data Collection 13 2.1 Study area ...... 13 2.1.1 Climate ...... 14 2.1.2 Soils ...... 16 2.1.3 Vegetation ...... 16 2.1.4 Higher trophic level ...... 18 2.2 Life history strategies of pikas ...... 19 2.3 Pikas found in the study area ...... 25 2.4 Grid trapping and observation ...... 26 2.4.1 Capture site and burrow characteristics ...... 26 2.4.2 Capture ...... 28 2.4.3 Observation ...... 30 2.4.4 Captured species and biomass of small mammals ...... 30 2.4.5 Welfare considerations in trapping and handling pikas ...... 32 2.4.6 Capture and observation success ...... 34 2.4.7 To trap or not to trap? Comparing trapping with observation . 35 2.5 Parasites ...... 40 2.6 Movement ...... 42 2.7 Summary ...... 46

3 Estimating Density 47 3.1 Introduction ...... 47 3.2 Methods ...... 48 3.3 Results ...... 51 3.4 Discussion ...... 52 3.5 Summary ...... 55

i CONTENTS

4 Life history 57 4.1 Age structure and life tables ...... 57 4.1.1 From weight to age ...... 58 4.1.2 Seasonal development of age structure ...... 69 4.1.3 Cohort life tables ...... 74 4.1.4 Summary ...... 78 4.2 Modelling survival rates ...... 79 4.2.1 Modelling background ...... 79 4.2.2 Survival rates ...... 81 4.2.3 Survival of early and late litter ...... 92 4.2.4 Summary ...... 94 4.3 Reproduction ...... 96 4.4 Population dynamics ...... 99 4.5 Discussing life history strategy ...... 101 4.6 Summary ...... 108

5 Ecosystem impact 109 5.1 Production and biomass removal ...... 109 5.2 Pika densities on regional scale ...... 124 5.3 Summary ...... 131

6 General Discussion 133 6.1 Life history strategy ...... 133 6.2 Ecosystem impact ...... 137 6.3 Further research needs ...... 141

References 143

Curriculum vitae 155

ii List of Figures

2.1 Location of the Gobi Gurvan Saikhan National Conservation Park . . . 14 2.2 Map of the study area comprising the Duund and Zuun Saikhan . . . . 14 2.3 Picture on the south facing slope of the Duund Saikhan ...... 15 2.4 Picture of the research camp ...... 16 2.5 Walther-Lieth diagrams of two climatic stations within the park . . . . 16 2.6 Gradient of soil proﬁles along an altitudinal transect ...... 17 2.7 Picture of burrow sampling plots ...... 18 2.8 Picture of the Mongolian and the Daurian pika ...... 21 2.9 Distribution of the Daurian pika ...... 21 2.10 Distribution of the Pallas pika ...... 22 2.11 Map of burrows and trap locations on the trapping site ...... 27 2.12 Mongolian pikas encountered by capture or observation ...... 36 2.13 Parasite load during the year ...... 41 2.14 Movement distances within two years ...... 43 2.15 Movement distances within and between sample sessions ...... 44

3.1 Density estimates ...... 53

4.1 Weight structure of captured pikas ...... 60 4.2 Average weight determined by growth curves ...... 62 4.3 Growth curve from maximum growth rates ...... 63 4.4 Discriminating juveniles from adults ...... 64 4.5 Predictions for age class composition ...... 65 4.6 Comparing growth of O. pallasi pricei, rufescens, curzoniae ...... 67 4.7 Cohort survival based on capture and observation ...... 70 4.8 Cohort survival based on capture ...... 70 4.9 Cohort survival based on capture for three summers ...... 71 4.10 Probabilities for two encounter histories ...... 79 4.11 Survival rates depending on density, sex, and age ...... 88 4.12 Survival rates for individuals from early and late litter ...... 95 4.13 Reproductive state during the year ...... 97 4.14 Trajectories for the numbers of males over females ...... 101 4.15 Two pikas ﬁghting over territories ...... 101

5.1 Hierarchical design of the exclosure experiment ...... 112 5.2 Estimates for standing crop based on habitat, time, and pika grazing . 115 5.3 Estimates for standing crop predicted by species group ...... 119 5.4 Pika burrow densities for diﬀerent livestock densities and altitudes . . . 127

iii LIST OF FIGURES

iv List of Tables

2.1 Precipitation in the summer months from 2000 to 2003 ...... 15 2.2 Standing crop along an altitudinal transect ...... 17 2.3 Rare and endangered animal species in the GGS ...... 18 2.4 Life history traits of Ochotona daurica and O. pallasi ...... 24 2.5 Habitat preference of Ochotona pallasi pricei and O. daurica ...... 26 2.6 Date, duration, and reference area of sample sessions ...... 29 2.7 Species captured on the trapping site ...... 31 2.8 Ochotona pallasi pricei captured and observed on the trapping site . . 34 2.9 Accuracy of identiﬁcation by capture and observation ...... 37 2.10 Length and regularity of encounters by capture and observation . . . . 38 2.11 Information gained by additional observation ...... 39 2.12 Information gained by additional capture ...... 40

3.1 Abundance and density predicted by diﬀerent estimators ...... 50 3.2 Movement distances based on capture and observation ...... 52

4.1 Median weights for all capture sessions ...... 61 4.2 Individuals used to construct a growth curve for maximum growth rate 63 4.3 Predicted cohort strengths based on winter and growth criteria . . . . . 64 4.4 Age structure at the end of the summers 2000-2002 ...... 71 4.5 Cohort Life Tables ...... 75 4.6 Life Tables for all sexes ...... 76 4.7 Life Tables for litter ...... 77 4.8 Propabilities to observe four diﬀerent encounter occasions ...... 80 4.9 Number of animals used for modelling ...... 82 4.10 Parameter coding for pooled data, using the most reliable information . 83 4.11 Parameter coding for capture data, spanning three summers ...... 84 4.12 Candidate models set ...... 85 4.13 Model likelihoods for recapture, pooled data ...... 86 4.14 Model likelihoods for survival, pooled data ...... 87 4.15 Model likelihoods for recapture, capture data ...... 89 4.16 Model likelihoods for survival, capture data ...... 90 4.17 Canditate models set for individuals from early and late litter . . . . . 93 4.18 Model likelihoods for early and late litter ...... 94 4.19 Pika life history traits revisited ...... 102

5.1 Number of pikas captured in the summers from 2001 to 2003 ...... 110 5.2 Plant species cover on sample plots ...... 111

v LIST OF TABLES

5.3 Precipitation in the summers of 2002 and 2003 ...... 111 5.4 Analysis of variance table for standing crop comparing two summers . . 114 5.5 Analysis of variance table for standing crop comparing species groups . 116 5.6 Signiﬁcant eﬀects from the ANOVA for species groups ...... 120 5.7 Production and biomass removal for a mountain steppe with burrows . 121 5.8 Ratio of burrow to steppe productivity along an altitudinal transect . . 125 5.9 Production and biomass removal along an altitudinal transect . . . . . 128

vi Acknowledgements

This project could never have been finished without the help of many people, friends and advisors and anything in between. Thank you for helping, fighting, growing, loving, enduring and encouraging me in so many various ways! I am grateful to Prof. Dr. G. Miehe, Prof. Dr. R. Samjaa, Prof. Dr. R. Brandl, Dr. V. Re- tzer, Dr. K. Wesche, Dr. J. Dauber, T. Monkhzul, R. Klug, Devlin, Carola, and Bettina for giving scientific advice and encouragement. V. Retzer, K. Wesche, S. Jansen, G. Miehe, V. Clausnitzer, T. Monkhzul, R. Undrakh, Enkhjargal, Bjambaa, Bekhee, S. Schmidt, Enkhshimek, Ojuntsetseg, Shurentsetseg, K. and E.-O. Nadrowski and many others gave invaluable help during the organisation of this project. Thank you Vroni, Vayu, Ria, Lou, Gouine, Frank, Sandra, Thomas, Christina, Tatjana, HikE, Monkhzul, Janis, Björn, Maike und Ines for your friendship and love. Thanks go to the staff of the Faculty of Geography of the Philipps-University of Mar- burg, the working-group of animal ecology in the Faculty of Biology of the University in Marburg, the staff of the Faculty of Biology of the Mongolian University, the staff of the GTZ-office in Mongolia, and the staff of the Research Center for Infectious Diseases in Dalandzadgad. The German Acadamic exchange, the German Science Foundation, and the Ministry of Economic Co-operation funded this project. Thank you Enkhjargal, Amgaa and K. Wesche for collecting plants and M. Pietsch for drying plants in Halle, and H. von Wehrden for supplying the maps. Thank you F. Behnsen for the drawing and pictogram of a Mongolian pika. I’m grateful to K. Wesche, V. Retzer, and my mother for reading through the last drafts, which greatly improved the present work. Thanks go to the open source community. This thesis was written using the open source softwares LATEX(www.latex-project.org) for typesetting, Xemacs (www.xemacs.org) as editor, R (www.r-project.org) for statistics and graphics, Xfig (www.xfig.org) for some of the figures, MARK (White and Burnham, 1999) for calculation of survival and recapture estimates. I am especially thankful to my daughter Janis, who was born shortly before this study was finished. You have brought sunshine in the winter months while I was finishing this study and I have learned so much from you already.

vii ABSTRACT

Life history strategy and ecosystem impact of a dominant small mammal herbivore in a mountain steppe Pikas (genus Ochotona) were suspected to possibly behave as “small mammal pests” in the mountain steppes of the Gobi Gurvan Saikhan National Conservation Park (GGS) in the Mongolian Gobi Altai. The park hosts rare and endangered animal species like the snow leopard (Uncia uncia), and the argali (Ovis ammon). Additionally it is an important rangeland for livestock of local herder families. The present study shows that the Mongolian pika (Ochotona pallasi pricei), a subspecies of the Pallas pika, is the dominant small mammal species in the mountain steppes of the GGS. It outnumbered other small mammals by one order of magnitude and outweighed them by two of them. This study presents the first account on trapping and observing individually marked Mongolian pikas. It therefore presents instructions on how to capture, handle, and observe individuals of this species. Data on capture and observation are available for three summers and one winter from a trapping grid of 100 100 m2, including a × summer of drought. Data based on observation was more reliable in terms of encounter success than data based on capture. Captured juveniles were discriminated from adults using information on the development of their weight. An average adult weight was estimated to range between 180 and 200 g. Possible scenarios for dynamics of population densities were simulated using a system of Leslie-matrices based on estimates for survival and reproduction of the observed individuals. Survival rates were estimated using maximum likelihood techniques with competing models for survival and recapture probabilities. The most parsimonious model included effects of population density, age, and sex on survival, while there was no effect of cohort affiliation nor of the climatic factors season and drought. However a classification of a summer and a winter season was useful to delimit age groups. Gener- ally, adults showed higher survival rates than juveniles, females showed higher survival rates than males, and survival declined with density. Estimates for reproduction were based on the observations of litter number and size, resulting in a median of 3 and a maximum of 13 juveniles per female pika. Although simulated population densities were similar to the measured population densities, they did not reflect the effects of the year of drought. Population densities were measured using pooled data from capture and observation sessions. In the study period pika densities varied between 14.6 and 49.8 individuals per ha. Densities were stable for most of the observed period of time, their interquartile range spanned 5.7 animals. Median density was 21.4 animals, which is less than the 28 burrows on the trapping site. Lowest densities were reached one year after the summer of drought, indicating a time lag of one year in the response to the drought conditions. Comparing reproductive effort, factors influencing survival rates, and density dynamics of the Mongolian pika with other species of the genus shows that this species exhibits traits of the group of non-burrowing pikas, which are closer to a K-type of life history strategy than the group of burrowing pikas. Density dependent survival indicates that burrow possession may be crucial for survival. Ecosystem impact of the pikas was assessed studying the productivity of the burrow habitat in comparison to steppe habitat together with the effect of grazing by pikas and by larger herbivores. Productivity and biomass removal was measured using exclosure plots on burrow and steppe habitat. Burrows showed higher productivity than steppe

viii ABSTRACT

habitats when water availability was higher. The grass Agropyron cristatum profited most from the burrow habitat in terms of biomass and quality. This grass is an important fodder plant for livestock. An effect of livestock grazing was missed by the experiment. The productivity of the vegetation was controlled by the availability of moisture, not by pika grazing, although individual pikas removed more biomass, when it was available. This biomass is probably stored in the burrows. Pikas preferentially grazed Agropyron cristatum on burrow plots. Burrows were estimated to last 120 years at least. The studied system behaved according to the prediction based on the non-equilibrium theory of rangeland dynamics: plant productivity was controlled by climatic conditions, as was the density of the herbivores. However, although pika densities varied, there were upper and lower limits to this variation made possible by the territorial behaviour of the animals. Possession of territories together with harvesting plant material enables the species to mitigate climatic inter-annual variability. Pika burrow densities were controlled by altitude and thus probably by the long- term availability of plant biomass. Livestock densities had only a small effect on burrow densities. However, this effect changed at the pediment angle separating pediments from the mountain ranges, indicating a change of system behaviour. The present study shows that the Mongolian pika cannot be seen as “small mammal pest” species, since its densities are controlled by the availability of burrows and it has a positive influence on pasture quality.

ix ZUSAMMENFASSUNG

Ub¨ erlebensstrategie und ok¨ osystemarer Einﬂuss eines dominierenden Klein- s¨augers in einer Bergsteppe

Kleinsäuger der Gattung Pfeifhasen (Ochotona) wurden verdächtigt, als “Plagen” in den Bergsteppen des Gobi Gurvan Saikhan Nationalparks im mongolischen Gobi Altai aufzutreten. Der Nationalpark beherbergt seltene und gefährdete Tierharten wie den Schneeleoparden (Uncia uncia) und den Argali (Ovis ammon). Zusätzlich wird er von den ansässigen Hirtenfamilien als Weideland genutzt. Die vorliegende Arbeit zeigt dass der Mongolische Pfeifhase (Ochotona pallasi pricei), eine Unterart des Pallas Pfeifhasen, die dominierende Kleinsäugerart in den Berg- steppen des GGS ist. Sie ub¨ ertrifft die anderen Kleinsäuger um eine Größenordnung in der Anzahl und um zwei Größenordnungen im Gewicht. Da die Arbeit erstmalig Fang- Wiederfang und Beobachtung individuell markierter Tiere des Mongolischen Pfeifhasen beschreibt, werden Anleitungen zum Fang, zur Handhabung und zur Beobachtung der Art gegeben. Ub¨ er den Zeitraum von drei Sommern und einem Winter wurden vom Jahr 2000 bis zum Jahr 2002 Informationen zu den markierten Tieren mithilfe von Fal- lenfang und Beobachtung auf einer 100 100 m2 großen Untersuchungsfläche gesam- melt. Der Zeitraum der Untersuchungen×schloss eine strenge Durre¨ im Sommer 2001 mit ein. Beobachtung erbrachte verlässlichere Informationen bezuglic¨ h des wieder- holten Antreffens markierter Individuen als der Fallenfang. Juvenile Tiere konnten aufgrund ihrer Gewichstentwicklung von Adulten unterschieden werden. Ein durch- schnittliches Adultgewicht wurde zwischen 180 und 200 g geschätzt. In einem Matrixmodell der Populationsdynamik wurden Ergebnisse zu Ub¨ erlebensraten und Reproduktionsraten der beobachteten Individuen zusammengefasst. Ub¨ erlebensraten wurden mit Maximum-Likelihood-Methoden ermittelt, wobei verschiedene Modelle mit Ub¨ erlebens- und Wiederfangraten miteinander konkurrierten. Das spar- samste Modell beinhaltete Populationsdichte, Alter, und Geschlecht als Einflussgrößen fur¨ Ub¨ erlebensraten, nicht aber die Zugehörigkeit zu einer bestimmten Kohorte oder die klimatischen Faktoren Saisonalität und Durre.¨ Allerding war Saisonalität in der Einteilung von Juvenilen und Adulten sinnvoll. Generell zeigten adulte Tiere höhere Ub¨ erlebensraten als juvenile, Weibchen höhere als Männchen und die Ub¨ erlebensrate sank mit der Populationsdichte. Reproduktionsraten wurden basierend auf Beobach- tungen zur Anzahl und Größe der Wurfe¨ geschätzt. Dies ergab einen Median von drei Juvenilen und ein Maximum von dreizehn Juvenilen pro Weibchen. Obwohl die simulierten und die gemessenen Populationsdichten vergleichbare Grö- ßenordnungen hatten, konnten das Matrixmodell die Auswirkungen der Durre¨ auf die Populationsdichte nicht abbilden. Populationsdichten wurden mithilfe der aus Fang und Beobachtung zusammengelegten Daten gemessen. Populationsdichten schwankten zwischen 14.6 und 49.8 Individuen pro ha. Mit einer Differenz von 5.7 Tieren zwischen oberen und unterem Quartal war die Dichte relativ stabil. Der Median der Dichten lag mit 21.4 Tieren unterhalb der Anzahl der Bauten auf der Untersuchungsfläche (28 Bauten pro ha). Die niedrigsten Dichten wurden im Jahr nach der Durre¨ gemessen. Dies macht einen Verzögerungseffekt der Durreb¨ edingungen um ein Jahr wahrschein- lich. Werden reproduktiver Einsatz, die Einflussgrößen der Ub¨ erlebensraten und die En- twicklung der Populationsdichten des Mongolischen Pfeifhasen verglichen mit dem an- derer Arten der Gattung, so zeigt diese Art Eigenschaften aus der Gruppe der “nicht- grabenden”-Pfeifhasen, die näher einem K-Typ von Ub¨ erlebensstrategien stehen. Die

x ZUSAMMENFASSUNG

Dichteabhängigkeit der Ub¨ erlebensraten legt nahe, dass der Besitz eines Baus entschei- dend fur¨ das Ub¨ erleben von Individuen ist. Ub¨ er die Produktivität der Bauten im Vergleich zur Steppe wurde der Einfluss der Pfeifhasen auf das Ok¨ osystem abgeschätzt. Gleichzeitig wurde die Beweidung durch Pfeifhasen und größere Herbivoren untersucht. Dazu wurde ein Ausschlussexperiment genutzt. Auf Bauten war die Produktivität dann höher, wenn auch mehr Feuchtigkeit vorhanden war. Das Gras Agropyron cristatum profitierte am meisten von den Bedingungen auf Bauten bezuglic¨ h Biomassezuwachs und Qualität der Biomasse. Eine Beweidung durch größere Herbivore konnte von diesem Experiment nicht nachgewiesen werden. Pfeif- hasen ernteten zwischen 10 und 35 % der vorhandenen Biomasse, wobei Agropyron cristatum auf den Bauten bevorzugt wurde. Der Biomassezuwachs der Vegetation wurde allerdings vom Niederschlag begrenzt, nicht durch den Effekt der Beweidung. Das untersuchte Ok¨ osystem bestätigte die Vorhersagen, die auf der Theorie einer Nichtgleichgewichtsdynamik von Weideland basieren: Sowohl die Produktivität der Vegetation als auch die Dichte der Herbivoren wurde von der klimatischen Variabilität kontrolliert. Allerdings variierten die Pfeifhasendichten nicht stark, es gab untere und obere Grenzen der Schwankungen. Diese Grenzen werden durch das territoriale Ver- halten der Tiere ermöglicht. Der Besitz eines Territoriums zusammen mit dem Ernten pflanzlicher Biomasse ermöglicht es der Art, die Auswirkungen der inter-annuellen klimatischen Variabilität zu mildern. Eine Bauten-Lebensdauer wurde auf mindestens 120 Jahren geschätzt. Bauten- dichten von Pfeifhasen wurden von der Höhenlage beeinflusst und damit wahrschein- lich von der Langzeitverfugbark¨ eit von Biomasse. Viehdichten dagegen haben nur einen kleinen Einfluss auf Bautendichten von Pfeifhasen. Da jedoch dieser Einfluss am Ub¨ ergang zwischen Pedimenten und Bergen sein Vorzeichen wechselt, können dort unterschiedliche System-Verhalten vorherrschen. Mongolische Pfeifhasen (Ochotona pallasi pricei) kommen somit nicht als “Klein- säugerplagen” in Betracht, da ihre Populationsdichten durch die Anzahl der Bauten begrenzt werden und sie die Weidequalität positiv beeinflussen.

xi ZUSAMMENFASSUNG

xii Chapter 1

Introduction

Small mammals are often seen as pest species, damaging crop, competing for forage or destructing pasture (Leirs, 1994; Stenseth, 1989), especially if they occur in high densities (Eldridge and Simpson, 2002). However, this is often based on emotional rather than scientiﬁc judgement (Putman, 1989). To assess whether a species can exhibit traits of a pest species, a sound knowledge of the biology of this species, especially the factors controlling survival and reproduction, is essential (Begon et al., 1996; Putman, 1989; Stenseth, 1989). Additionally, small mammals usually are an integral part of ecosystem dynamics (Jones et al., 1997; Kinlaw, 1999). For this reason, the repercussions of reduced densities on ecosystem functioning should be considered before deciding to control a focus species (Bourli`ere, 1975). Aim of the present study is to gain insight into the life history of a dominant small mammal in a mountain steppe, which has been considered a small mammal pest species. Special attention is given to the factors controlling the survival of this species. Additionally, the study analyses its impact on the mountain steppe in terms of grazing and burrowing activity.

1.1 Mountain steppe

Grasslands are fascinating ecosystems. The virtual absence of trees enables to see vast distances and oﬀers aesthetic and spiritual gratiﬁcation. One of the attractions of grasslands are large herds of wild ungulates like bisons or wild horses. But they also form the basis of livelihood for the people living in them. Many of our food grains, as for example wheat or millet, originated in grasslands (White et al., 2000). About 40 % of the globe’s terrestrial areas are grasslands in the wider sense (White et al., 2000). Steppes in a more formal sense are temperate natural grasslands, characterised by dry summers and cold winters, and a vegetation consisting of xeromorphic grasses, perennial herbs and low dwarf-shrubs (Coupland, 1992a). Steppes are characterised by extremely variable moisture and temperature conditions. Ecosystems with consistently more precipitation host forests, while those with consistently less precipitation host deserts. Plants in steppe environments are tolerant to a wider range of temperature and moisture conditions than typical desert or forest species (Lavrenko et al., 1993). Steppes in Eurasia extend from eastern Europe, along the Black Sea and Kaza- khstan, to Mongolia and China (Lavrenko et al., 1993). They can be characterised by two major gradients: a hygric gradient from north to south and a gradient of continen-

1 INTRODUCTION

tality from west to east. Variability of moisture and temperature conditions increases along the gradients of aridity and continentality. Productivity of all trophic levels is higher in the north-western parts than in the south-eastern parts of this vast region. The continental steppes east of the Ural mountains are characterised by mountainous landscapes, expressing a well deﬁned altitudinal diﬀerentiation of vegetation. Here, precipitation rises with altitude so that mountains within otherwise desert environments support steppes. These mountain steppes are more isolated and soils are not as deep as in steppes in the more westerly and northerly areas (Lavrenko et al., 1993).

1.2 Herbivores

Organisms of higher trophic levels, which depend on plant productivity, have to cope with the variability of climate intrinsic to steppe ecosystems. Under the more temperate and moist conditions in the North-Western part of the Eurasian steppes, where plant productivity is less variable, they can maintain relatively constant densities and the systems can sustain complex trophic relationships. Under the more variable conditions in the South-Eastern part, organisms of higher trophic levels have to cope with the rapid response of plant productivity to varying climatic conditions, so that trophic relationships are less complex (Lavrenko et al., 1993). However, herbivory is intrinsic to steppe dynamics (Lavrenko et al., 1993; Wesche and Retzer, 2005). Herbi- vores and steppe plants have undergone a common evolutionary history. The dominant life form of bunch grasses in steppes is not only adapted to climatic variability, but also to grazing and trampling. Adaptations include storage organs below ground and their ability to reproduce by means of below ground tillers (Lavrenko et al., 1993). Using space below ground is also common to the animals found in these ecosystems. Formozov (1968) described four types of strategies for mammals in Eurasian steppes: (a) non-hibernating burrowers such as the Brandt’s vole (Microtus brandtii), the steppe lemming (Lagurus lagurus) or the Daurian pika (Ochotona daurica), (b) hibernating burrowers such as the souslik (Citellus pygmaeus) and the marmot (Marmota bobac), (c) geophiles (mammals living most of their times below the soil surface), such as the zokor (Myospalax myospalax), and (d) gregarious, far roaming ungulates such as the khulan (Equuus hemonius) or the Mongolian gazelle (Procapra gutturosa). The three first strategies include species using burrow systems. Burrows provide protection from environmental extremes, lessen risk of predation, and allow access to hoarded food (Kinlaw, 1999). Hoarding food as done by the burrowers, or moving far distances as done by the far roaming wild ungulates, represent key resources (Illius and Connor, 1999) for the animals, mitigating the effects of adverse climatic conditions. Additionally, large herbivores are better suited to digest fibre than smaller herbivores (Clauss and Hummel, 2005), while smaller mammals are generally held to be the better competitors due to their ability to graze at lower plant heights (Farnsworth et al., 2002). Animals have developed digestive strategies to use the plants rich in fibre which are common in steppe ecosystems. By rumination, ungulates hold the plants back in the rumen, so that microbes can develop, which are then digested by the ruminants. The same strategy is used by horses and lagomorphs, although in them the microbes are situated in the blind gut (caecum; Wehner and Gehring, 1990).

2 INTRODUCTION

1.3 Life history strategies

An organism’s life history is its lifetime pattern of growth (including death), differen- tiation, storage, and reproduction. The study of life histories is a search for patterns, a search for associations between one life-history trait and another, or between life- history traits and features of the habitats in which the life histories are found. The rules that have been found, describing when particular traits might be favoured, work within constraints: the possession of one life-history trait may limit the possible range of some other trait, and the general morphology of an organism may limit the possible range of all its life-history traits. Life histories are fixed within limits by the genotype of the individual, the present and past environmental conditions, and the interactions between the two (Begon et al., 1996). Body weight is perhaps one of the most apparent aspects of an organism’s life history. It puts constraints on other life history components. For example, terrestrial warm-blooded vertebrates show a bimodal distribution of body weights (Bourlière, 1975). Large mammals tend to live longer, reproduce later, and produce less offspring than small mammals. They tend to follow an iteroparous strategy, producing offspring in a series of separate events, during and after each of which the organisms maintain themselves in a condition that favours survival to reproduce again. Relative to this small mammals tend to live a more semelparous strategy, producing all of their offspring over one relatively short period in a single reproductive event and not investing in subsequent survival. Generally speaking, the small mammal strategy of exploitation of environmental resources is based upon a rapid adaption to short-term changes in population and/or environmental modifications. This is also an example illustrating the influential concept of r and K selection. The letters refer to parameters of the logistic equation, where the intrinsic rate of natural increase (r) is only realised in the absence of competition. The per capita rate of increase of a population is reduced to zero growth when a population reaches the capacity (K) of the respective environment. The logistic equation leads to a sigmoidal increase of population density with time. The “r/K concept” envisages two contrasting types of individuals (or species). It has emerged from the contrast between species that were good at rapidly colonising “empty” islands (r-type) and species that were good at maintaining themselves on islands in a later stage of succession (MacArthur and Wilson, 1967) and was then applied more generally (Boyce, 1984; Pianka, 1970). In this context, r-selected individuals have been favoured for their ability to reproduce rapidly, whilst K-selected individuals have been favoured for their ability to make a large proportional contribution to a population which remains at its carrying capacity (K). While K-selected populations mostly live in ecosystems that experience only small environmental fluctuations, r-strategists should prosper in systems which are either unpredictable in time or short-lived. The predicted traits for an K-type of life history are higher adult survival rates, larger size, larger and fewer offspring with more parental care. Population densities are fairly constant and competition between individuals is strong. In contrast, predicted traits for an r-type of life history are a larger reproductive allocation of resources. Here adult survival is predicted to be lower and there are more and smaller offspring compared to K-strategists. Actual survival and population densities are expected to vary considerably depending on variation in the abiotic conditions of the environment (Begon et al., 1996). Although this dichotomy is an oversimplification (Stearns, 1977, 1992), the concept

3 INTRODUCTION

has been influential and thought-provoking (Begon et al., 1996). It has been useful in describing some general differences among taxa and within taxa. As for the present study, the concept is used in a heuristic manner to illustrate two possible patterns of life history strategies, one resulting in stable densities (K-strategy) through density- dependent survival rates, and one resulting in fluctuating densities, where survival rates are not primarily controlled by density (r-strategy). Generally, a small mammal in a dry steppe environment would be expected to show patterns of life history traits conforming more with an r-strategy because of the small body size and the unpredictable environment. Examples which confirm these patterns in respect to size within Eurasian steppes are the Brandt’s vole (Microtus brandtii, Samjaa et al., 2000) or the Plateau pika (Ochotona curzoniae, Smith et al., 1990). Both species show varying population densities and high reproductive rates. In contrast, the Asian wild ass, or khulan, (Equuus hemonius, Reading et al., 2001) is a large sized ungulate which exhibits K-type life history traits, such as long life and low reproductive rates.

1.4 From keystone species to small mammal pest: Ecosystem impacts

Burrowing small mammals are considered as a prime example of ecosystem engineering (Jones et al., 1997; Scott Mills et al., 1993). Ecosystem engineers are organisms that directly or indirectly modulate the availability of resources to other species, by causing physical state changes in biotic or unbiotic materials (Jones et al., 1994). Kinlaw (1999) reviews impacts of small mammal burrowers on geomorphology, pedological effects, impacts on vegetation effects and the animal community. Burrows generally have higher organic soil content and are more productive than the surrounding vegetation (Butler, 1995; Holtmeier, 1999; Tsendzhav, 1985; Wesche et al., submitted). An example of this is Brandt’s vole Microtus brandtii, which can destroy pasture by burrowing and foraging (Samjaa et al., 2000), but this pasture will carry more productive and nutrient rich plants after the vole populations have left the area (Weiner and Gorecki, 1982 quoted in Holtmeier, 1999). Thus, through their burrowing activity, small mammals indirectly influence the vegetation. However, they do so also directly. Through grazing they can alter vegetation structures such as height, cover, and vegetation composition (Huntly, 1987; Tsendzhav, 1985). Several species of burrowers have been classified as keystone species (Power and et al, 1996), because the variety of impacts of burrowers in arid ecosystems makes their importance disproportionate to their abundance (Kinlaw, 1999). The classification as keystone species also refers to their function for higher trophic levels. Especially non-hibernating small mammals serve as prey base for mammalian and avian predators (Schaller, 2000; Smith et al., 1990). Human exploitation introduces a new perspective on burrowing small mammals. Today’s steppes are no longer in their natural state but completely used by humans. Most of the former steppe ecosystems have experienced heavy conversion to agricul- ture, which have destroyed the original steppe (Lavrenko et al., 1993; White et al., 2000). Most of the remaining steppes are used as rangeland for nomadic pastoralists (White et al., 2000). From the human perspective system components can now be put

4 INTRODUCTION into value. Economic “loss” caused by wild animals can be calculated and they can be seen as causing “damage”. Former keystone-species and ecosystem engineers can now be called “small mammal pest”, exactly because they have a strong impact on the ecosystem (Allaby, 1999). But the classification of a species as “pest” often is problematic. The origin of the word “pest” is the epidemic disease bubonic plague, it’s German name being “Pest”. This word stands for a catastrophic disease which in its worst pandemic for Europe during the years 1347 to 1352 killed a third of the human population, leading to thoroughgoing changes in the economy and culture of that time (Brockhaus, 1999). Ironically, the species of focus in this thesis is not only considered to be a possible “small mammal pest”, but is also a vector for the disease bubonic plague (Smith et al., 1990), which still exists in Mongolia (Ebright et al., 2003; National Research Centre for Infectious Diseases, pers. comm.). Small mammals are generally known to be vectors for diseases harmful for humans (Barnett and Dutton, 1995; Grunell and Flowerdew, 1990) and rats are known as vectors for plague. At the same time, mouse-like animals arouse emotions of fear and disgust in Europeans, where mice feature cartoons with women standing on chairs and screaming when seeing a mouse. In the fairy-tale of the rat-catcher, a musician is promised the wealth of a city if he manages to remove the rats. Since considering animals to be “pest” species often involves emotional judgement, it is difficult to decide whether control actions are scientifically sound or if emotional judgement overestimates the “damage” caused by small mammals. (Putman, 1989). The question, if herbivores can “destruct” an ecosystem is directly linked to the question, which factors control ecosystem dynamics. This has been addressed by different lines of thought, namely systems ecology and rangeland theory. In systems ecology Hairston et al. (1960) first took the approach of analysing ecosystems as food chains, representing trophic levels as if they were populations. They proposed that herbivores had little impact on the dominant plant species, because they were limited by their predators, not their food supply. In their exploitation ecosystem model Oksanen et al. (1981) proposed that trophic structure varies with primary productivity; more productive communities are able to support longer food chains. Thus, the three-trophic-level ecosystem with plants, herbivores and carnivores becomes a special case. The exploitation ecosystem model predicts that the importance of herbivory varies with primary productivity because of this change in food chain length: in extremely unproductive habitats, resident herbivore populations are not supported by the low phytomass, and plants are resource-limited. As productivity increases, herbivore populations can be supported and the vegetation becomes limited by herbivory. Further increases in productivity allow carnivores to persist, producing the three-trophic-level scenario. Yet even higher productivity might support a fourth trophic level, resulting in herbivores again limiting vegetational biomass. The exploitation ecosystem model, although developed in terrestrial systems, seems to have been particularly successful in predicting dynamics of aquatic ecosystems (Huntly, 1991). Rangeland theory differentiates between equilibrium and non-equilibrium systems (Illius and Connor, 1999; Wiens, 1984). The traditional so called “equilibrium theory” of rangeland science assumes that climatic conditions are essentially stable and that vegetation is in a largely stable equilibrium with the climate. Under such conditions, livestock is the main ecological factor controlling ecosystem changes in grasslands. Being so dominant, livestock is also held responsible for being the driving force behind pasture degradation (Fernandez-Gimenez and Allen-Diaz, 1999). In contrast,

5 INTRODUCTION

the “non-equilibrium” theory emphasises the overwhelmingly important role played by abiotic conditions and especially by the highly variable precipitation patterns. Non-equilibrium conditions are widely assumed to be very common in semi-arid to arid regions including steppes, where water is the limiting factor for plant growth. As variability of precipitation often increases with increasing aridity of the climate (Sullivan and Rohde, 2002), inter annual changes in precipitation patterns are particularly strong in drylands. Thus, both plant and livestock communities can be expected to experience major fluctuations in numbers reflecting the variability of the moisture conditions. Generally, non-equilibrium characteristics are becoming progressively more important with increasing dryness of the sites (Wiens, 1984), while equilibrium conditions and widespread overgrazing phenomena are typical for environments with relatively reliable precipitation. Several thresholds have been proposed, but most authors expect that environments with less than 200 mm annual precipitation and a coeffi- cient of variation larger than 30 % in inter annual precipitation are better described by non-equilibrium models (Fernandez-Gimenez and Allen-Diaz, 1999; Illius and Connor, 1999).

Thus, although small mammals are widely seen as harmful for pasture conditions (Zhong et al., 1985), this judgement is not always scientifically sound. In cases where small mammals did cause economic harm, often human action made this possible in the first place. For example, the catastrophic changes in the ecosystems in Australia caused by the European rabbit (Oryctolagus cuniculus) were caused by human introduction of this species (Holtmeier, 1999). Human made monocultures, like corn fields, can also experience loss by foraging small mammals (Leirs, 1994). Even the damage caused by the native Brandt’s vole in the Mongolian steppes have been seen in connection with overgrazing by livestock (Samjaa et al., 2000; Smith et al., 1990; Zhong et al., 1985). Additionally, non-equilibrium theory predicts that damage caused by higher trophic levels is less probable in dry ecosystems (see above).

To decide, whether control action should be taken, a good knowledge of the density dynamics of the focus species and the ecosystem is crucial (Begon et al., 1996; Putman, 1989). Even if a species is considered a pest, often a laissez-faire approach is less costly than implementing controls. To decide upon costs of control, repercussions of reduced densities on ecosystem functioning should be considered. If control is decided upon, knowledge on the biology of the focus species is crucial for eﬃcient control. As an example, it is not eﬃcient to reduce density of a small mammal showing cycles in population dynamics, when population density is high. Such measures have been shown to maintain the conditions for high reproduction. It is better to reduce densities in the low phases in order to prolong them (Leirs, 1994).

Even without knowledge of system behaviour and the biology of the focus species, small mammals are generally seen as pest species. They are controlled in Asia, including Mongolia, by distributing poisons harmful for all mammals, including humans (Smith and Foggin, 1999; Smith et al., 1990). This is even seen as relevant for foreign aid. One of the aims of a poverty reduction loan by the International Foundation for Agricultural Development was “eradicating rodents”. This was thought to be a means to improve pasture conditions (UB Post, 7.11.02, 16.1.03).

6 INTRODUCTION

1.5 Situation in Mongolia and project background

The vast grasslands of Mongolia cover 1.3 million m2 and offer some of the world’s finest examples of extensively used central Asian steppes (Lavrenko et al., 1993; Sneath, 1998; White et al., 2000). However, Mongolian steppes are by no means pristine environments, as nomadic pastoralism has been the principal land use practice for centuries (Fernandez-Gimenez and Allen-Diaz, 1999). Although nomadic pastoralism is widely held to be the most sustainable land use form (Retzer, 2004; White et al., 2000), overgrazing is still an issue in Mongolia (Fernandez-Gimenez and Allen-Diaz, 1999; Hilbig and Opp, 2005). With the change to market economy political changes since 1989 have had tremendous effects on the country’s economy, and subsequently on the grazing systems in the region, as pastoral land use is still the main source of income (Fernandez-Gimenez and Allen-Diaz, 1999; Retzer, 2004). The end of planned economy implied growing unemployment and in- security of income. Together with a rapid population growth, many families could no longer make a living in the urban centres. This resulted in an increasing number of herder families and even stronger increases in the number of livestock (Muller¨ , 1999). The change of the political and economic system from 1989 onwards coincided with a suit of climatically favourable years, with enough precipitation to sustain more livestock (Retzer, 2004). On a nationwide scale, livestock numbers remained essentially stable at about 24 million heads throughout the larger part of the twentieth century. However, after 1989 they showed some 30 % increase (Retzer, 2004). This was accompanied by phenomena of overgrazing and degradation especially along roads and water sources in all parts of Mongolia (Hilbig and Opp, 2005). The strong concern about the rising numbers of livestock during the 1990’s was tragically fulfilled by a sequence of droughts and zuuds from 2000 to 2003 (Retzer, 2004). Zuuds are events of winter snow, where livestock cannot reach the vegetation anymore. Livestock numbers were reduced again to their longtime average of about 24 million heads (Retzer, 2004). In the winters of 2000 and 2001 pictures of piles of starved livestock carcasses in Mongolia went around the world. With the change to market economy, the Mongolian government put strong ef- forts into nature conservation. From 1994, when the Protected Areas Law was passed, to 1997 the amount of land with protected area designation increased up to 11 % (Bedunah and Schmidt, 2000). One of six National Conservation Parks designated by the Ministry of Nature and Environment in 1994, the Gobi Gurvan Saikhan National Conservation Park, hosts desert to mountain steppe vegetation (Bedunah and Schmidt, 2000; Wesche et al., submitted). The project “Nature conservation and buffer zone development” of the GTZ (German Technical Assistance) aids in monitoring and managing the park resources including the perspectives of the local herder families (Bedunah and Schmidt, 2000). In view of the rising numbers of livestock during the 1990’s there was a growing concern about degradation and desertification of the natural resources of the country (Miehe, 1996). Since livestock roams free in Mongo- lia, protected areas are always additionally rangelands (Bedunah and Schmidt, 2000; Reading et al., 1999; Retzer, 2004), and rising livestock numbers can thus lead to degra- dations even in protected areas. In order to map the vegetation including forms and indicators of degradation, and to propose areas for special protection, Sabine Miehe visited the Gobi Gurvan Saikhan National Conservation Park. Although she did not find simple indicators for degra-

7 INTRODUCTION dation, she could show that the whole area is under high grazing pressure. She additionally was concerned about the apparent connection between small mammal and livestock density (Miehe, 1996, 1998). Georg Miehe then initiated a memorandum of understanding between the department of geography, Philipps-University of Marburg, and the biological faculty of the Mongolian State University, Ulaanbaatar, Mongolia, which was signed in 1999. By 2000, the German Science Foundation (DFG), the Ger- man Academic Exchange Program (DAAD), and the German Ministry for Economic Co-operation and Development (BMZ) granted funds to start research projects and found a research station in the National Conservation Park with the aim to study the carrying capacity of the mountain-steppe biocenosis with respect to the transformation of nomadic pastoralism. The question, if the dominant small mammals in the mountain steppes of the park represent “pests” is of central importance within this project. It was not known, which of the two pika species occurring in the area, the Daurian and the Pallas pika (Ochotona daurica, O. pallasi), was the most abundant small mammal, nor is much of the biology of either species known (Smith et al., 1990).

1.6 Aim of this study

The present study aims at contributing data and scientiﬁc information as a basis for the discussion of the presence of “small mammal pests” in the mountain steppes of the GGS. To assess the need of controlling the species and possible repercussions on the ecosystem after control, it aims at sketching possible scenarios for population dynamics of the focus species and estimating their impact on the standing crop by grazing and burrow construction. To achieve this, the following questions had to be answered:

Which is the most dominant small mammal species in the mountain steppes of the GGS? Pikas are the most abundant small mammals in the mountain steppes of the park (Reading et al., 1999), but it was not known, which of the two species to expect, Ochotona daurica, or Ochotona pallasi.

– To assess, which of the species dominates the mountain steppes, samples were taken from diﬀerent biotopes within the research area (Section 2.3, p. 25), while

– the dominance in term of biomass was estimated by trapping and weighing small mammals on a representative steppe site (Section 2.4.4, p. 30).

Can this species be expected to reach high population densities? To sketch possible scenarios for population dynamics, the factors controlling survival and reproduction, as well as emigration and immigration have to be studied. Pikas don’t disperse far (Smith, 1987), so that immigration and emigration were not the main focus of this study. Nonetheless, movement distances were measured from marked individuals (Section 2.6, p. 42). However, scenarios for population dynamics may be anticipated by studying the life history traits of the focus species. The genus Ochotona

8 INTRODUCTION shows a dichotomous pattern in life history traits similar to the r/K-concept, in which one group of species tends to have constant population densities, while the other group shows ﬂuctuating densities (Section 2.2, p. 23, and Smith, 1988; Smith et al., 1990). Thus, the following questions had to be asked:

(a) Which population densities can be observed at present?

(b) Which factors control survival?

(d) What type of population dynamic can be expected using estimates for survival and reproduction?

(e) Which type of life history strategy does the focus species show?

(a) Measuring population densities Present pika densities were measured using information from capture and observation sessions (Chapter 3, p. 47).

(b) Which factors control survival rates? The two life history strategies described for pikas suggest that survival rates are either mainly inﬂuenced by climatic conditions or by population density (Section 2.2, p. 23 and Smith, 1988; Smith et al., 1990). Therefore, the inﬂuence of

– intra- and inter-annual variation in climate and

– population density on the survival of the diﬀerent age- and sex-groups were to be studied.

Preliminaries for studying survival Before analysing survival rates, the following questions had to be answered:

– How can Mongolian pikas be captured and observed? Since the present study is the ﬁrst to have captured and recaptured pikas in Mongolia, optimal methods for trapping, marking, and observation had to be found (Section 2.4, p. 26). The present study explicitly payed attention to the welfare of the individuals trapped and handled (Section 2.4.5, p. 32).

– How can age be determined? Since age is one of the most common factors controlling survival rates in living organisms, age groups had to be distinguished. The present study used weight development to discriminate age groups (Section 4.1.1, p. 58).

9 INTRODUCTION

Measuring survival To study the factors controlling survival rates for the diﬀerent animals, individuals were marked and observed during a whole year, including the winter, and an additional summer. Survival rates were then measured using – life tables (Section 4.1.3, p. 74) and – capture-recapture modelling based on maximum-likelihood techniques (Section 4.2.2, p. 81). – As indicator of individual health, parasite load was measured (Section 2.5, p. 40).

(c) What is the reproductive capacity of the focus species? Litter size and number were measured together with signs of reproductive activity in Section 4.3 on page 96.

(d) What type of population dynamic can be expected using estimates for survival and reproduction? Based on a system of Leslie-matrices the dynamics of population density resulting from the above analysis of survival and reproduction was compared to the measured population densities on the study site (Section 4.4, p. 99).

(e) Which type of life history strategy does the focus species show? Con- sidering all the above information the life history strategy of the studied population can be discussed (Section 4.5, p. 101).

Can the focus species be expected to degrade the pasture? Apart from removing biomass small mammals effect pasture quality by their burrowing activity. Their beneficial or detrimental effect on the pasture is then dependent on the area influenced by burrowing. If pika density is additionally facilitated by livestock grazing, their effect on the ecosystem may by amplified by human activity. The present study therefore asks: (a) How is standing crop influenced by burrow and steppe productivity and by biomass removal? (b) How intense is burrowing activity of the pikas? (c) Are burrow densities effected by livestock densities?

(a) Burrow and steppe productivity and biomass removal The following questions were addressed using a replicated exclosure experiment on burrow and steppe habitat (Section 5.1, p. 109): – In how far do pika burrows as special habitats improve/damage pasture quality in terms of standing crop and plant species performance? – How much do pikas remove from the standing crop? – How much does livestock remove from the standing crop? – Is this inﬂuenced by diﬀerent amounts of precipitation?

10 INTRODUCTION

(b) Burrowing activity Burrows were mapped and changes are observed on the mapped study site over the research period from 2000 to 2002 (Section 2.4.1, p. 26).

(c) Burrow number and livestock density The effect of livestock density on pika burrow density for different altitudes was studied using line-transects at different altitudes in the Duund Saikhan and Zuun Saikhan mountain ranges (Section 5.2, p. 124).

11 1.6. AIM OF THIS STUDY INTRODUCTION

12 Chapter 2

Basic Data Collection

This chapter introduces the study area with special attention given to pika burrows. It further introduces the objects of the study, the pikas (Section 2.2, p. 19). This chapter additionally answers questions that form the basis of the study of life history patterns and ecosystem impact later on. First of all it clariﬁes, which is the dominant pika species in the study area (Sec- tion 2.3, p. 25). To study life history pattern of a species, factors that control survival and reproduction for single individuals have to be observed. For this it is necessary to identify individuals. This chapter therefore asks how pikas can be captured, marked, and observed, so that information on individuals can be gained. To assess how this can be done with the least interference with the individuals’ lives, welfare considerations are included (Section 2.4, p. 26). This section also includes basic information on capture success and estimates by what factor the pikas are dominant compared to other small mammals in terms of numbers and biomass. Regarding factors important to the survival, parasite load during the year is studied as an indicator of the health of the observed population (Section 2.5, p. 40). As to the question, if pika population can rapidly spread to distant areas, mobility of individuals is studied in Section 2.6 on page 42.

2.1 Study area

The Gobi Gurvan Saikhan National Conservation Park (henceforth GGS) is part of the Gobi Altai. As part of the Altai mountain ranges, the Gobi Altai is the south- easternmost outcrop with the lowest altitudes. The highest peak of the GGS is the Zuun Saikhan which reaches 2815 m above sea level. The formation of the mountain chain started in the Tertiary and the area is still tectonically active (Barthel, 1990). Major landforms within the park are mountain ranges, pediments and inter-montane basins. Pediments are linearly eroded by gullies, which are only temporarily water-ﬁlled. The pediment angle, which separates the plain pediment area from the mountains, is found at about 2350 m in the Duund Saikhan. Within the mountains, steep rocky slopes form a heterogenous landscape with high relief energy (Retzer, 2004; Wesche et al., in press). Politically, the park is situated in the Omnogobi aimag, west of its capital Dalandzadgad (Figure 2.1, p. 14). The name “Gobi Gurvan Saikhan” translates as “Three Beauties of the Gobi” and refers to the mountain ranges Baruun Saikhan, Duund Saikhan and Zuun Saikhan

13 2.1. STUDY AREA 2. BASIC DATA COLLECTION

Ulaanbaatar

Gobi Gurvan Saikhan Figure 2.1: Location of the National Conservation Park Gobi Gurvan Saikhan Na- Dalandzadgad tional Conservation Park in Mongolia. Draft H. von Wehrden.

1500 m asl Duund Saikhan

Research camp

2000 m asl Zuun Saikhan

Site 2 Yoliin Am

Figure 2.2: Map of the study area with the locations of the research camp and the sites visited for collecting pika individuals within the Duund and Zuun Saikhan mountain ranges. Diﬀerent shadings of grey for every 100 m diﬀerence in altitude. Draft H. von Wehrden.

(translating “Western, Middle, and Eastern Beauty”). These mountain ranges are found in the very east of the National Conservation Park. The study area comprises

the Duund Saikhan and the Zuun Saikhan (Figure 2.2, p. 14), while most of the data

¡ ¡ was collected at the research site (approx. 43 36.949 N, 103 46.450 E) on the upper pediments at the south facing slope of the Duund Saikhan (Figures 2.3, p. 15 and 2.4, p. 16).

2.1.1 Climate The region is characterised by a highly continental climate with precipitation occurring mainly in the summer months. Variability of precipitation is high in time and space. Within the park area two moisture gradients can be detected. Precipitation increases from west to east and with altitude. Long-term precipitation records are

available from a governmental weather station some 30 km SW of the research station. ¢ Bayandalai (1570 m asl) reports a mean annual temperature of 4.5 ¢ (January -18 ,

14 2. BASIC DATA COLLECTION 2.1. STUDY AREA

Figure 2.3: Picture on the south facing slope of the Duund Saikhan from 2000 m asl showing the transition from gently sloping pediments to ragged mountain relief. The gers of the research camp can be seen as a white spot on the foot of the mountains in the right part of the picture.

July 20 ¢ ), while mean precipitation is 110 mm with a high coeﬃcient of variation (around 35%; data from National Meteorological Service, quoted from Retzer, 2004). The aimag capital Dalandzadgad, 80 km NE of the research station is located 100 m lower and shows an annual mean of 131 mm precipitation. Short term precipitation records at the research station are available for a full year in 2000/2001 and for the year 2003/2004. In all years, precipitation was clearly higher than in Bayandalai, suggesting that precipitation at the research station ranges between 130 and 160 mm. It can be assumed that the average amount of precipitation at the research camp is relatively similar to that measured in Dalandzadgad. However, the upper slopes receive higher precipitation (Retzer, 2004).

The summer of 2001 brought a severe drought in the whole of southern Mongolia. Precipitation at the Table 2.1: Precipitation research site in July and August was only 16 % of the at the research site in the amount received the year before. In the summer of 2002 Duund Saikhan mountain in the summer months the situation was similarly severe, although precipitation from 2000 to 2003. Data was almost twice as much as during the same period in from Retzer (2004) and 2001 (Retzer, 2004). In contrast to these two years, pre- Wesche and Retzer (2005). cipitation in July and August of 2003 was 53 % higher than in 2000 (Table 2.1, p. 15, Wesche et al., submitted). Date Rain (mm) Fernandez-Gimenez and Allen-Diaz (1999) suggested a 2000 64.2 threshold of 200 mm annual precipitation and a coeﬃ- 2001 35.2 cient of variation larger than 30 % in interannual pre- 2002 24.4 cipitation separating non-equilibrium from equilibrium 2003 94.6 systems. This threshold might have been reached in the mountains, while the pediments would still clearly experience non-equilibrium conditions.

15 2.1. STUDY AREA 2. BASIC DATA COLLECTION

Figure 2.4: The three gers (Mongolian felt-tent) of the research camp located at 2350 m asl. In the front of the picture fences and a small observation tent adjacent to the trapping site.

Dalandzadgad Bayandalai

Figure 2.5: Walther-Lieth diagrams of the two climatic stations within the park nearest to the research camp, Bayandalai and Dalandzadgad. Data from the Me- teorological Service Mongolia, draft K. Wesche.

2.1.2 Soils While Burozems are the dominant soil types in the pediment region, Kastanozems prevail on the southern side of the mountain ranges and Chernozems and Paracher- nozems on the northern side. At the research site on the south facing slope of the Duund Saikhan the Burozems get better developed from the intermontane basins towards the pediment angle and change from a light towards a dark Burozem. In the mountain ranges the Burozems are replaced by Kastanozems which change from a light to a dark version with increasing altitude (Figure 2.6, p. 17, Retzer, 2004; Wesche and von Wehrden, 2002, T. Hennig, pers. comm.). Soil conditions under pika burrows diﬀer from the surrounding steppe. The burrows are enriched in nutrients, whereas water availability is lower (Wesche et al., submitted).

2.1.3 Vegetation The GGS National Conservation Park comprises desert to mountain steppe vegetation, including patches of forest in the Zuun Saikhan. Vegetation at the research station is governed by communities intermediate between the relatively moist steppes of the up-

16 2. BASIC DATA COLLECTION 2.1. STUDY AREA

B U R O S E M B U R O S E M K A S T A N O S E M K A S T A N O S E M K A S T A N O S E M ( P E T R I C C A L C I S O L ) ( P E T R I C C A L C I S O L ) ( P A L C I C K A S T A N O Z E M ) ( H A P L I C K A S T O N O Z E M ) ( P H A E O Z E M ) 2 0 0 0 m 2 2 0 0 m 2 3 0 0 m 2 6 0 0 m 2 8 0 0 m

c m g r a v e l c m g r a v e l c m g r a v e l c m c m 0 0 0 0 0 O d Q Q Q r o c k ( d e b r i s ) d e b r i s

1 0 A k h c 1 0 A k h 1 0 A k h 1 0 1 0 A x h 1

2 0 2 0 2 0 2 0 2 0 B s - C c A h A x h

3 0 3 0 3 0 3 0 3 0 I I B s - C c C m c A x h 2 4 0 4 0 4 0 4 0

l C c l C 5 0 5 0 5 0 5 0

I I I f A h c - B t 6 0 6 0 6 0 6 0 f A h - i l C

I I C m c 7 0 7 0 I I i l C n I V l C c I I i C m c

8 0 b e d r o c k 1 . H 2 0 2 . H 2 2 2 . O r 1 6 . H 2 6 2 . H 2 8 Figure 2.6: Gradient of soil profiles along the altitudinal transect in the Duund Saikhan. Data from T. Hennig, draft by K. Wesche. per mountain slopes and the dry desert steppes of the lower pediments (Wesche et al., in press). Species from the mountain steppe comprise Agropyron cristatum, Stipa krylovii and Artemisia frigida, while desert steppe species are Stipa gobica and Al- lium polyrrhizum (Hilbig, 1995; Wesche et al., in press). Mountain slopes host three main groups of plant communities, namely mountain steppes, dominance stands of the dwarf shrub Artemisia santolinifolia, and scrub composed of Juniperus sabina (Wesche and Ronnenberg, 2004). Productivity increases along the altitudinal gradient from the pediments to the mountainous environments (Table 2.2, Retzer et al., in press). On the upper pediments and in the mountain steppes, pika burrows form an easily recognisable special habi- Table 2.2: Standing crop tat. They host different vegetation and show an earlier along an altitudinal transect phenology (Retzer, 2004). Two principal burrow types at the research site in the Du- und Saikhan. Data are from can be clearly differentiated within the surroundings of Retzer et al. (in press). They the research camp by the dominance of dwarf shrups were collected in the drought (mainly Artemisia santolinifolia), or the dominance of summer of 2001. grasses (mainly Agropyron cristatum, Figure 2.7). The latter are very common on the slightly sloping pediments, Altitude Standing crop while those with A. santolinifolia are restricted to special m asl kg ha-1 sites along erosion gullies or at the foot of scree slopes, 2000 43 where soil movement is always present even in the ab- 2200 43 sence of pikas (Wesche et al., in press). Burrows occupy 2400 260 7-12 % of the surface in the surroundings of the research 2600 200 station (Retzer, 2004). With a median size of 38.1 m2 2800 320 (Section 2.4.1, p. 26) they are large structures, suggesting that they last at least several decades (Whitford and Kay, 1999). Indeed, during

17 2.1. STUDY AREA 2. BASIC DATA COLLECTION

Figure 2.7: Sampling plots on a burrow with and without access for pikas. The picture on the left shows a sampling plot without fence. The grass Agropyron cristatum was grazed, while Allium spec. was not. To the right a sampling plot which was fenced. In this plot Agropyron cristatum grew considerably higher.

Table 2.3: Examples of rare and endangered animal species in the GGS. This list is by no means comprehensive! Conservation status according to CITES-convention, the Mongolian Red Book (MRB), and the Mongolian hunting law (MHL). Summarised from Reading et al. (1999).

English name Scientiﬁc name CITES MRB MHL

Goitred Gazelle Gazella subgutturosa * * Khulan (Wild Ass) Equus hemonius * * * Ibex Capra sibirica * * Argali (Wild Sheep) Ovis ammon * * * Snow Leopard Uncia uncia * * Saker Falcon Falco cherrug * Peregrine Falcon Falco peregrinus * Merlin Falco columbarius * Lesser Kestrel Falco naumanni * Houbara Bustard Chlamydotis undulata * * * Eagle Owl Bubo bubo * four years of observation, we did not observe a single new burrow being built, although the present burrows were constantly being changed and modiﬁed by the pikas (Section 2.4.1, p. 26).

2.1.4 Higher trophic level The Gobi Gurvan Saikhan National Conservation Park hosts an extraordinary wealth of large wild herbivores, including regionally and globally endangered species such as the black-tailed gazelle, the khulan (Reading et al., 2001) or Bactrian camels. On the upper pediments and in the mountainous environments, the ibex is considered rare by the Mongolian hunting law and the argali is endangered in Mongolia as well as

18 2. BASIC DATA COLLECTION 2.2. LIFE HISTORY STRATEGIES OF PIKAS globally. Several predator species, including the globally threatened snow leopard, and several rare and endangered avian raptors are found within this area (Table 2.3, p. 18, Reading et al., 1999). Small mammals found on upper pediments and mountainous environments in the park are the grey ratlike hamster (Cricetulus migratorius), the Mongolian gerbil (Meriones unguiculatus), the Daurian pika Ochotona daurica, or the Pallas pika Ochotona pallasi. Their status and distribution in the park is hardly studied (Reading et al., 1999). Although the National Conservation Park is protected by law since 1994 and the mountainous areas are part of the special zone where human land use should be prohib- ited, virtually the entire region is under permanent and complete anthropo-zoogenic inﬂuence (Bedunah and Schmidt, 2000; Reading, 1996; Stumpp et al., 2005). Live- stock having been observed at the research site included all the species traditionally herded in Mongolia, which are goats, sheep, cattle (including yaks), camels, and horses. 95 % of the animals using the region around the research camp from October 2000 to October 2001, including areas within the special zones of protection, are livestock (Retzer et al., in press).

2.2 Life history strategies of pikas

The dominant small mammals in the mountain steppes of the GGS are pikas (genus Ochotona). Reading et al. (1999) states that the Pallas pika Ochotona pallasi and the Daurian pika Ochotona daurica are well distributed throughout the park. Pikas are plague vectors (National Research Centre for Infectious Diseases, pers. comm.). In regular intervals they are monitored for the presence of the disease in the park area. Pikas are remarkably homogeneous in general morphology and body mass, although independent studies showed that major adaptive themes within the genus have evolved several times (Smith et al., 1990; Yu et al., 1997). Pikas are a poorly known group of mammals, most species occur almost exclusively in remote settings. Most of the pikas are restricted to Asia, where 23 species are found. Pikas are widely distributed in central Asia from the Tibetan plateau to the dry desert steppes of the Gobi and form a notable and important component of the fauna wherever they occur in the Northern Hemisphere (Smith et al., 1990). Living representatives of Ochotona form a monotypic genus within the family Ochotonidae (order Lagomorpha), which was clearly differentiated from the other lagomorphs as early as the Oligocene (Dawson, 1967). The genus probably originated in Asia followed by a widespread dispersal (Dawson, 1967; Thenius, 1980). It reached its peak of diversity in the Miocene, when Ochotonids extended throughout the Palaearctic and into Africa. By the Pleistocene Ochotona was found in the eastern United States and as far west in Europe as Great Britain. This extensive spread was followed by restriction to its present range in Asia and North America. The Mongolian name for pikas, “ukher oghdoy”, may translate as “cattle-like mouse”, stressing the difference in size between pikas and rodents occurring sympatri- cally. It may also reflect the different foraging strategy of pikas compared to rodents: pikas, like all lagomorphs, are hindgut fermentors, reingesting their food and digest- ing it in the blind gut (caecum; Wehner and Gehring, 1990). This form of digestion is similar to ruminants, which also reingest their food. Since cattle are ruminants, the association “ukher oghdoy” of cattle and pikas is suitable. The Latin genus name

19 2.2. LIFE HISTORY STRATEGIES OF PIKAS 2. BASIC DATA COLLECTION

“Ochotona” goes back to the Mongolian word “oghdoy”, although this is generally used for all small mammals in Mongolia. Smith et al. (1990) review the importance of pikas in ecosystems in terms of gener- ating and maintaining biodiversity due to their role in the food-web, as herbivores, and as ecosystem engineers. They found that pikas form an important prey base for mammalian and avian predators, either as preferred food, or as buffer species (Smith et al., 1990; Tsendzhav, 1985; Zevegmid, 1975). This is even more important during winter, since pikas do not hibernate. As an additional ecosystem impact, they directly influence the population dynamics of other herbivores by their typical caching behaviours. Other small mammals, native and domestic ungulates, may accumulate in areas with their hay piles, especially in winter (Naumov, 1974). Pikas further provide shelter for other species of birds and mammals (Smith and Foggin, 1999). Moreover, burrowing pikas are prime excavators and thus qualify as allogenic engineers (Kinlaw, 1999). They contribute positively to ecosystem-level dynamics by recycling and mixing soil. They have been shown to increase local productivity and plant species richness (Tsendzhav, 1985; Wesche et al., submitted). Pikas also directly influence vegetation composition by grazing (Huntly, 1987; Tsendzhav, 1985) There is an apparent relationship between grazing of domestic animals and pika population densities. Black-lipped pikas (O. curzoniae) may be found at greater densities when the height of vegetation has been reduced by domestic grazing (Zhong et al., 1985). Daurian pikas (O. daurica) were more likely to contribute to the deterioration of rangelands that were already overgrazed (Shi, 1983). Higher population densities of pikas at places with higher livestock densities were observed in the study region by Sabine and Georg Miehe (Miehe, 1996, 1998) and were one of the reasons to initiate this research project. There has been a wide-scale defaunation of vertebrates in Asia as the result of control measures directed at burrowing species of Ochotona. The most serious ecosystem- level problem concerning the control of lagomorphs involves black-lipped, Daurian, and Pallas‘s pikas (Ochotona curzoniae, O. daurica, O. pallasi) in China. Control began in 1958, reached a peak between 1963 and 1965, and continues on a reduced scale (Smith et al., 1990). There is a continuous emphasis on control programs from the highest levels of administration in China (Smith and Foggin, 1999; Smith et al., 1990). In Mongolia, control of small mammals using poison is also funded by the government (R. Samjaa, pers. comm.). Workers of the park administration of the GGS mentioned considerable concern about increasing pika densities after the cessation of poisoning with the change to market economy (Miehe, 1996). Their density is reduced by shooting individuals at highly frequented tourist sites by the National Research Centre for Infectious Diseases, although the effect of this measure stays questionable. Herders interviewed by V. Retzer did not mention concern about pikas (pers. comm.), although they did so in interviews conducted by S. Miehe (pers. comm.). In the study region, two pika species occur: Ochotona daurica (Pallas 1776) and O. pallasi (Gray 1867). O. daurica has longer claws, its’ toe pads are covered with fur (Zevegmid, 1975), and does not have a rust-coloured patch at the throat, while Ochotona p. pricei has shorter claws, bare toe pads and a rust-coloured patch at the throat (Smith et al., 1990; Zevegmid, 1975). Figure 2.8 on the facing page shows the two species. The two species are among the most drought-tolerant pika species, with O. daurica being widespread in Mongolian grass steppes, while O. pallasi is found in dry mountain steppes and desert steppes of the Mongolian and Gobi Altai (Smith et al.,

20 2. BASIC DATA COLLECTION 2.2. LIFE HISTORY STRATEGIES OF PIKAS

Figure 2.8: To the left the head of a Daurian pika (Ochotona daurica) taken by V. Retzer, to the right a Mongolian pika (Ochotona pallasi pricei) with hair dyed orange by picric acid. The Picture was taken by M. Pietsch.

Figure 2.9: Distribution of the Daurian pika (Ochotona daurica) in Central Asia. Map from Smith et al. (1990), changed.

1990; Sokolov and Orlov, 1980). Both species have been considered as “pests” and have been subject to control eﬀorts for their potential competition with domestic livestock on open rangelands or for serving as vector for plague in Mongolia, China, Tuva, and Kazakhstan. The Daurian pika is distributed throughout Mongolia, extending to the desert steppes of China (Figures 2.9, p. 21 and 2.10, p. 22). It is listed as an abundant species in the status summary of Smith et al. (1990) and inhabits semi-desert and steppe environments. Much of its ecology appears similar to the closely related black-lipped

21 2.2. LIFE HISTORY STRATEGIES OF PIKAS 2. BASIC DATA COLLECTION

Figure 2.10: Distribution of the Pallas pika (Ochotona pallasi) in Central Asia. Map from Smith et al. (1990), changed. pika (O. curzoniae), which is well studied compared to the other Asian pikas (Smith, 1988; Smith and Foggin, 1999; Smith et al., 1986; Wang and Wang, 1996). O. daurica and O. curzoniae frequently have been regarded as conspecific, later, however, they have been treated as sibling species (Smith et al., 1990). Using molecular techniques, Yu et al. (1997) showed that the species might not be as closely related as morphological and behavioural characteristics suggest. Similarities between both species include high but fluctuating densities, showing a mostly monogamous mating system with fre- quent expressions of affiliative behaviour within family groups. Adults have three basic vocalisations, an alarm call, a long call, and shorter trills. The long call is a song-like trill that slows down in frequency toward the end. The shorter trills are similar, but not as loud and much shorter. Long calls and trills can be used to separate O. daurica from O. p. pricei in the field, since the latter only utters alarm calls. The distribution of the Pallas pika is restricted to the Altai, stretching from Kaza- khstan, Tuva, Russia, Mongolia to China (Figure 2.10, p. 22, Smith et al., 1990). It is listed as abundant but possibly locally threatened by isolation in the status summary of Smith et al. (1990). The range of the Pallas pika is more disjunct than in other species of burrowing pikas. The population in Kazakhstan is isolated from the population in the Mongolian Altai, and populations within the Mongolian Altai are again isolated from each other. Three subspecies are currently widely recognised: O. p. pallasi (Gray 1867)in northeastern Kazakhstan, O. p. pricei (Thomas 1911a)in Mongolia, Tuva, and China, along the Mongolian border, and O. p. hamica (Thomas 1912b) in China and perhaps southern Mongolia. O. p. hamica is poorly known. The trenchant differences in ecology and behaviour between the forms of O. p. pallasi and O. p. pricei suggest the possibility that they may be specifically distinct (Smith et al.,

22 2. BASIC DATA COLLECTION 2.2. LIFE HISTORY STRATEGIES OF PIKAS

1990). The subspecies occurring in the study area is O. p. pricei, here called the Mon- golian pika, as practised by Russian and Mongolian literature (Demin, 1962; Krylova, 1973; Rothshild and Derevshchikov, 1994; Sokolov and Orlov, 1980; Zevegmid, 1975; Zonov et al., 1983). All three subspecies have been documented to decline over the last centuries especially in the southern part of their range , which has been thought to be an eﬀect of the global drying of climate in the southern Gobi over the past 200 years (Kniazev and Savinetski, 1988; Kniazev et al., 1986 quoted in Smith et al., 1990, p.41). Ochotona p. pricei occupies arid steppe environments where it may either burrow in open steppe or live in rocks. The mating system is primarily polygynous, with males demonstrating high levels of intrasexual aggression. This subspecies appears to be more aggressive than O. p. pallasi, as aggressive relationships are ten times more fre- quent (Proskurina et al., 1985). Vocalisations are also diﬀerent between O. p. pallasi and O. p. pricei: the former utters a song or long call and soft quiet trills in meetings with females, while the latter does not (Smith et al., 1990).

Pika life history strategies Most pikas show one of two different life history patterns, which seem to be associated with the habitat the species occupies (Smith, 1988; Smith et al., 1990). Talus habitats are rocky areas, consisting of big stones. The pikas occupying talus-habitats adopt a K-type of life history strategy, they show constant population densities and relatively low fecundity rates (Section 1.3, p. 3). These species do not dig their own burrows but are dependent on the stony habitats that provide nesting sites. An r- type of life history strategy is adopted by all other species, occurring mostly in steppe habitats, but also in forests or shrub-dominated environments. These species dig their own burrows. Although pikas in general show striking similarities in morphology and behaviour, burrowing species can be distinguished by shorter vibrissae and more powerful and straight claws. These morphological traits seem to have evolved several times (Smith et al., 1990; Yu et al., 1997). There are few intermediate species, burrowing pikas occupying talus or rock habitat, but also occurring in steppe habitat. Demographic parameters, population densities, and social behaviour are substantially different between the two predominant patterns of life histories (Smith, 1988). Species of non-burrowing pikas are relatively asocial, comparatively long-lived, and have relatively stable low population densities and fecundity rates. In contrast, burrowing species normally are highly social, short-lived, may have high but fluctuating population densities, and have high fecundity rates (Table 2.4 on the next page). In a detailed interspecific comparison between the burrowing Ochotona curzoniae (closely related to O. daurica) and the non-burrowing O. princeps, Smith (1988) showed that survival rates differ in the two species. While juvenile survival was higher in the burrowing species, it was even higher for those juveniles born later in the summer. Smith suggested the close proximity of grazing sites as a reason for higher survival. Pikas in steppe habitats can spend more time feeding than pikas living in talus-habitats, where feeding sites are further from their shelter. Thus, burrowing pikas do not appear to be limited by forage during most of the April–July reproductive season. In contrast, the non-burrowing species showed low juvenile survival, and especially low survival in juveniles born later in the year. Adult survival was high in the non-burrowing species, even overwinter survival. This resulted in a lower ratio of juveniles and yearlings in the populations of the non-burrowing species. Additionally, survival was density-dependent in

23 2.2. LIFE HISTORY STRATEGIES OF PIKAS 2. BASIC DATA COLLECTION

Table 2.4: Life history traits within the genus Ochotona and for Ochotona daurica and O. pallasi, subspecies pallasi and pricei, as reviewed by Smith (1988) and Smith et al. (1990). The species occurring in the study area are O. daurica and O. p. pricei. Both species tend to be closer to the r-type life history pattern of burrowing pikas. Question marks indicate not yet studied life history traits.

Burrowing Ochotona Life history traits no yes daurica p. pallasi p. pricei

Reproduction Litters 1–3 2–5 3 2-3 3 Litter size 1–7 1–13 1–11 mean 8 1–12 In their summer of birth no yes yes yes yes Survival Juveniles low high ? ? ? Late litter lower higher ? ? ? Adults high low ? ? ? Overwinter high low ? ? ? Density dependent yes no ? ? ? Age structure Old age 6 2 4 ? 4 % Yearlings low high high ? high Density dynamics Density (/ha) 2.6–30 0–380 0.1–300 >100 ? Density variation low high high high high Density regulation territories climatea climatea ? ? competition

a Extreme climatic events (flooding, snow, cold, drought) and high overwinter mortality. the non-burrowing species, only individuals holding a territory could survive. The patterns of life history traits coincided with the variability of environmental conditions. Steppe environments, where burrowing species occur, are variable in precipitation and forage availability. In contrast, talus environments in mountainous regions show less variability (Smith, 1988). Intraspecific variation in demographic parameters showed that even non-burrowing species could exhibit higher fecundity rates and less stable population densities in environments having less predictable environmental conditions (Smith, 1988). Thus, Smith et al. (1990) and Smith (1988) propose a trade-off for all pika species between high investment in present survival and high investment in present reproduction triggered by habitat quality. Like in it’s close relative Ochotona curzoniae, populations widely fluctuate in size in O. daurica (Smith et al., 1990; Zevegmid, 1975). This fluctuation has been attributed to flooding, snowless winters, summer droughts, rain during autumn (which causes stored plants in hay piles to rot), and to competition with other grazing herbivores (Table 2.4). Tsendzhav (1976) reported densities varying between 25 and 55 (mean 29) individuals at 12 different sites in the Mongolian Khangai. Young pikas may make up 93 % of the population (Shubin), but other studies reported far lower percentages and even life-spans of up to 4 years (Tsendzhav, 1977). At the same time the reproductive

24 2. BASIC DATA COLLECTION 2.3. PIKAS FOUND IN THE STUDY AREA

rate is high. Litter size ranges from one to eleven, each female is capable of producing several litters per year and juveniles may reproduce in the same summer in which they were born (Zevegmid, 1975). Ochotona daurica is reported to be a monogamous species (Smith et al., 1990). Tsendzhav (1976) reported a mean of 15.5 burrows per hectare with a mean of 29 pikas (n=12). Population densities may also vary greatly in O. pallasi, although no numbers are available for Ochotona p. pricei (Table 2.4). The Mongolian pika appears to be relatively long-lived. Based on cement lines in the periosteal zone of the mandible Krylova (1973) determined survival rates of 50 % per year and a maximum longevity of four years. Two-year-old animals comprised 20 % of the summer population. On the other hand Shubin reported up to 90 % juveniles at the end of July. Additionally, the fecundity rate of both subspecies is high. There are no detailed data on the survival of age groups within both species available, and only anecdotal information on factors inﬂuencing population dynamics. From the data available, both species would be expected to show an r-type pattern of life history traits, like expressed in O. curzoniae, although both species seem to get comparatively old compared to other burrowing species.

2.3 Pikas found in the study area

The ﬁrst obvious question that has to be answered is, which species is the more abundant in the mountain steppe of the study area? Of the two pika species reported (Ochotona daurica and Ochotona pallasi pricei, Reading, 1996), habitat preferences are not known. Smirnov (1974) stated that in case of close proximity to steppe pikas (O. pusilla), O. p. pallasi selects diﬀerent habitats, namely hill slopes and rocky habitats. It is thus expected that the most common pika on steppe habitats is Ochotona daurica.

Methods In a joint study with the Mongolian Institute for the Research of Infectious

Diseases, pikas were collected by shooting in diﬀerent habitats in the Zuun Saikhan at

two locations (approximately 43.4851 N, 104.0793 E and 43.4530 N, 103.9768 E, Fig- ure 2.2, p. 14). This shooting is a routine exploration to search for infectious diseases, none of the animals was shot solely for the purpose of studying habitat preference. Small mammals are shot all over Mongolia by the Institute. Habitats were chosen to reﬂect the mountain steppe community, but water surplus places were additionally visited, since Ochotona daurica was heard preferentially in these places beforehand. Habitat classiﬁcation was based on Wesche et al. (in press). Habitats chosen were (a) Juniperus sabina communities, (b) Mountain steppes, (c) Gully within mountain steppes, (d) Kobresia mats on upper slopes, (e) Saline meadows, (f) Semi-permanent river beds. Altitudes ranged between 2300 and 2400 m asl.

Results At 21 diﬀerent sites 54 individuals of Ochotona pallasi pricei and 9 of Ocho- tona daurica were collected. Table 2.5 shows the distribution of the species between the habitats. Ochotona p. pricei was the most abundant species in hall habitats, except those with permanent water supply. These were the sole habitats where Ochotona daurica was found.

25 2.4. GRID TRAPPING AND OBSERVATION 2. BASIC DATA COLLECTION

Table 2.5: Total number of individuals collected during a joint excursion with the Institute for the Research of Infectious Diseases in diﬀerent habitats. Habitat classiﬁcation was based on Wesche et al. (in press).

Habitat Ochotona Visited sites Altitude [m asl] p. pricei daurica min max

Juniperus sabina communities 5 0 1 2400 2400 Mountain steppes 16 0 7 2330 2390 Gully within mountain steppes 4 0 2 2390 2390 Kobresia mats on upper slopes 2 5 4 2370 2370 Saline meadows 0 4 1 2255 2255 Semi-permanent river beds 27 0 6 2300 2377

Discussion Ochotona pallasi pricei prevails in the mountain steppes of the study area, while O. daurica occurs only on moister sites. These findings are in accordance with Ognev (1940), who noted that the species occurs in fertile parts of the Gobi Desert and in low-lying places, which may be subject to flooding. Proskurina et al. (1985) noted that in comparison to Ochotona pallasi pricei the Daurian pika inhabits optimal habitats with abundant food. Ochotona daurica has been shown to exhibit strong fluctuations in population density (Smith et al., 1990 and references therein) and is target of control measures in China (Zhong et al., 1985). At the same time, the observation of the park management that pika densities was increasing was made at a site with permanent water supply. It was made in Yoliin Am, a highly frequented tourist site, where a small permanent stream is the attraction. The ice developing at this site in winter stays until the summer months and sometimes even until the next winter. The park management was concerned about the safety of tourists, since pikas carry the plague (pers. comm.). These concerns and observations may have been transferred to the whole park area and to all pika species occurring in the park.

2.4 Grid trapping and observation

This section introduces the methods used for capture and observing the focus species. It then asks, to what extent the Mongolian pika dominates the small mammal community in mountain steppes in terms of number and biomass (p. 30). After having established, that this species is by far the most dominant small mammal in the ecosystem, all the following questions only refer to the Mongolian pika. Since this study is the ﬁrst to use capture-recapture and observations of individually marked animals of this species, special attention is given to the welfare of individuals during capture and handling (p. 32), to the capture techniques (p. 34), as well as to the comparison of encounter and identiﬁcation success of capture and observation (p. 35).

2.4.1 Capture site and burrow characteristics A representative area of one hectare comprising both steppe and erosion gully near the research station was chosen as trapping site (Figure 2.11). 38 burrows were mapped on

26 2. BASIC DATA COLLECTION 2.4. GRID TRAPPING AND OBSERVATION

Erosion gully Location of traps

120 Slope Pediment steppe 100 80 60 40 Distance in meter 20 0 −20 −20 0 20 40 60 80 100

Distance in meter

Figure 2.11: Map of burrows and trap locations on the trapping site. At each trap location a stone pile was erected to facilitate orientation during observation sessions. The trapping site comprised gully and pediment habitat.

this area and adjacent to it, 27 of which fell into the trapping site, two further burrows were about half way within the site (Burrows with the coordinates (0,0) and (0,35) in Figure 2.11). This amounts to about 28 burrows on one hectare. The area of the single burrows ranged between 2.5 and 71.9 m2. Median burrow size was 38.1 m2, while ﬁrst and third quartile were 24.3 and 52.7 m2 respectively. This corresponds to circles with diameters of 6 to 8 m with a median at 7 m. Nearest neighbour distances from the burrow centroids to each other had a median of 14 m for the ﬁrst and 17 m for the second nearest neighbour. The observed Mongolian pikas did not dig much. During the three years of observation, there was no change in the location and size of the burrows. Pikas were only observed to repair their entrances. Based on these observations and on data from

27 2.4. GRID TRAPPING AND OBSERVATION 2. BASIC DATA COLLECTION opening three burrows at the research site (unpublished data), the following comments are given on burrow development: since pikas were constantly repairing their burrows, some centimetres of new tunnels may emerge each year. If a yearly increase in tunnels of 20 cm is assumed; if further each burrow may consist of four complete tunnels, a burrow with median area (38.1 m2, 6.2 m diameter for a circular burrow) would take 124 years to develop. No single individual would be able to dig its own burrow. This is in accordance with the ﬁndings of Whitford and Kay (1999), who state that large burrows last for a long time. The burrows are probably much older, since they are passed from one generation to the next. Nonetheless, Lavrenko et al. (1993) classify pika burrows as non-permanent structures in comparison to marmot Marmota bobac burrows. At the same time, pikas readily accepted new objects put in their vicinity to sit on them and dig under them. Here, tunnel development was much faster than on established burrows. On the trapping site (Figure 2.11), there were several smaller burrows, which were not occupied by one pika individual, e.g. the burrows at the coordinates (85/65), (19,9), or (52,8). If climatic conditions get better for a longer time, these small burrows may develop into larger sized ones, thus increasing burrow density. On the other hand, observations of abandoned burrows in the Sevrej mountains (own observation) and the Nemegt mountains (V. Clausnitzer, pers. communication), might indicate that climatic conditions were better in the past.

2.4.2 Capture Capture started in July 2001 on a 60 60 m2 site comprising pediment steppe and × gully habitat (Coordinates (10,0) to (70,60) in Figure 2.11). The trapping grid was arranged in an 7 7 arrangement, trap intervals were 10 m, with one trap placed at the edges of the ×grid cells. In September 2001, the grid was enlarged to an 11 11 × arrangement on 100 100 m2 (Table 2.6 on the next page and Figure 2.11). Traps were placed at places×frequented by pikas, not necessarily exactly at the edges of grid cells. Edges of grid cells were marked with piles of stone. Capture took place in nearly monthly intervals (Table 2.6). The capture session in March 2001 was interrupted by heavy storms. 41-54 Extra Large (10 12 38 cm) and the 67-80 Large (8 9 23 cm) × × × × Sherman live traps were used (www.shermantraps.com). Animals were identified by using Sokolov and Orlov (1980). Different baits were offered in the traps, including palatable fresh (cabbage) and dry (dried plants) and unpalatable material (wool). Wool was carried away to the burrows, but neither of the baits increased capture success. Morrison and Teitz (1956) showed that captive Ochotona collaris preferred carrots and hay, but did not touch peanut butter or corn, while Kawamichi and Liu (1990) successfully baited Ochotona curzoniae with apples and flowers. Capture followed a robust design with primary and secondary sample sessions (Krebs, 1999). In this design, one primary sample session consists of several secondary sample session. In the following, if “sample session” is used without specification, it refers to primary sample sessions. Since O. pallasi followed a diurnal activity pattern and so did not enter traps in the night and additionally seemed not to have orientation in the dark, traps were set in the night before and checked for the first time around 10:00 in the morning, followed by successive checking during the rest of the day. During the day traps were controlled

28 2. BASIC DATA COLLECTION 2.4. GRID TRAPPING AND OBSERVATION

Table 2.6: Average dates, time diﬀerence in days between ﬁrst and last sampling date, and the size of the study site for the 16 sample sessions. Combining capture and observation led to increasing length of sampling time, since observation started several days after capture beginning in May 2001. This was done because hair dyed with picric acid needed some days to develop the brightest colour.

Only capture Capture and observation Study site Nr. Average date Time interval [d] Average date Time interval [d] [m m] × 1 06.07.2000 3.70 60 60 × 2 17.07.2000 0.95 60 60 × 3 23.08.2000 0.96 60 60 × 4 28.09.2000 0.17 100 100 × 5 26.10.2000 1.33 100 100 × 6 28.11.2000 1.35 28.11.2000 1.39 100 100 × 7 23.01.2001 1.22 23.01.2001 1.22 100 100 × 8 23.03.2001 6.00 100 100 × 9 25.04.2001 1.47 25.04.2001 1.47 100 100 × 10 22.05.2001 1.92 26.05.2001 9.56 100 100 × 11 24.06.2001 1.72 28.06.2001 9.17 100 100 × 12 27.07.2001 1.54 30.07.2001 8.12 100 100 × 13 22.08.2001 1.88 25.08.2001 7.88 100 100 × 14 23.09.2001 2.68 25.09.2001 7.01 100 100 × 15 01.07.2002 4.31 02.07.2002 6.13 100 100 × 16 07.07.2002 1.00 11.07.2002 8.29 100 100 ×

at least every 3 hours to prevent heat stress (see p. 32). Morning sessions were thus usually longer than the sampling sessions in the rest of the day . Handling was done outside the trapping site in the first months, until the car was used as mobile handling and observation laboratory from November 2000 on. Using the car greatly reduced the effort, since handling and writing was not hampered by wind or precipitation anymore. Traps with animals inside were transported to the car and placed in the shadow of the car while waiting for handling, to avoid heat stress. Successful traps were not replaced during handling, but after handling of all captured animals. If rodents were captured instead of pikas, food was offered to them. No food (fresh or dry plant material, bread) was accepted by pikas. Animals were weighed, sexed, marked, and inspected for parasites and bruises. Although some animals were weighed a second time within one primary sample session to assess weight loss caused by trapping, most of the animals were not handled when captured again and immediately released instead. Animals were marked using ear tags and dying the fur with picric acid. Ear tags had 6 different colours for front and back side and individual numbers (Mini-Ohrmarken, Tierzuchtgeräte-Strietzel, Gun¨ thersleben, Germany). They had to be shortened to reduce weight. Picric acid was applied symmetrically on the fur of the animal on head, neck, front side, middle side or back side. The dyed fur turned to bright orange after two days. Tattooing the ear, ear cuttings, and hair cuttings were additionally tried for individual marking, but these methods did not work well. For tattooing, pika ears were too hairy, ear cuttings following the patterns proposed by Kawamichi and Liu (1990) could only be located after some months, but forms

29 2.4. GRID TRAPPING AND OBSERVATION 2. BASIC DATA COLLECTION

could not be distinguished anymore. Kawamichi and Liu (1990) reported that ear clipping in adult Ochotona curzoniae and O. cansus worked for up to 98 days, but sharp outlines were lost when clipping young individuals. Patterns cut in the hair (Barnett and Dutton, 1995; Grunell and Flowerdew, 1990) were only visible for about a month during summer. During winter this method was not tested. It was assumed that an intact fur coat is important to protect the animals from the cold. In June, traps were put in higher densities at selected burrows to capture and mark more juveniles. Further information was gained by occasional ﬁndings of dead pikas or remainders of pikas (for example, one time an ear tag next to the guts of a pika).

2.4.3 Observation To facilitate observation on the trapping site, the stone piles at the grid corners were painted in two different colours. Those having 0, 50, or 100 in either X- or Y-coordinate were painted blue, dividing the site into four quarters, while all other stone piles were painted in yellow (Figure 2.11, p. 27). Systematic observation of the whole study site started in November 2000 from within a car placed at the middle of the study site (Table 2.6, p. 29). It had windows to all sides and the whole study site could be observed using binoculars with 10 magnification. In November 2000 and January × and April 2001 observation took place during the capture sessions. From May 2001 onwards, observations were conducted at least two days after capture which allowed the dyed hair to turn into its brightest colour. In March 2001 observation was aborted. Probably caused by heavy storms, no pikas could be observed on the study site for several subsequent days. In September 2001 and July 2002 a larger area was observed. For this the car was placed at different locations to scan the study site plus a border of about 25 m. Within 2 to 6 hours the whole study site was scanned until no new individual could be detected any more. Animals observed within and adjacent to the study site were noted with their coordinates. Location was estimated in reference to the distance to the stone piles constructed at the edges of the trapping site (see Figure 2.11, p. 27) or by using the mapped location of the burrows. Care was taken not to count unmarked animals twice within one observation session. This was done by assuming that pika movement was centred around burrows during the observation sessions. Unmarked animals observed simultaneously on neighbouring burrows were not counted as new observations.

2.4.4 Captured species and biomass of small mammals To determine which species dominates the small mammal community in the mountain steppes, the number of individuals captured during the 16 capture session (Table 2.6, p. 29) is listed and the biomass of each species within mountain steppes is approximated. Typical adult pika weight is 190 g, as will be shown later in this study (p. 62). Weight of the other species was calculated as average of all ﬁrst captures of these animals. Data from other capture events (not from the trapping site) were included for Meriones unguiculatus.

Results During the capture sessions 166 individuals of 4 species could be identiﬁed. A total of 155 belonged to Ochotona pallasi pricei, 7 of them were Cricetulus migra-

30 2. BASIC DATA COLLECTION 2.4. GRID TRAPPING AND OBSERVATION

Table 2.7: Number of individuals cap- Session cm m op oi tured in unbaited traps during cap- 1 2000.07.06 0 0 15 0 ture sessions for the four species en- 2 2000.07.17 0 0 11 0 countered: Cricetulus migratorius (cm), 3 2000.08.23 0 0 8 0 Meriones unguiculatus (m), Ochotona 4 2000.09.28 0 0 14 0 pallasi (op), and Oenanthe isabellina 5 2000.10.26 1 0 16 0 (oi). Note that O. isabellina is a bird. 6 2000.11.28 0 0 15 0 7 2001.01.23 0 0 7 0 8 2001.03.23 2 0 10 0 9 2001.04.25 0 0 11 0 10 2001.05.22 1 0 18 0 11 2001.06.24 2 0 62 1 12 2001.07.27 2 0 31 1 13 2001.08.22 3 1 18 0 14 2001.09.23 1 0 14 0 15 2002.07.01 0 0 22 1 16 2002.07.07 0 0 17 0 Sum 12 1 289 3

torius, a rodent, 3 of them Oenanthe isabellina, a bird species, and one of them was a Meriones unguiculatus, another rodent. 7 individuals of Cricetulus migratorius and 13 individuals of Meriones unguiculatus could be used to calculate an average weight for these species. Weight of C. migratorius ranged from 24 to 48 g, with a median at 36 g. Weight of M. unguiculatus ranged from 40 to 64 g with a median at 52 g. Table 2.7 shows the frequencies of the four species for the 16 sampling sessions. A χ2 test did not find significant differences in the proportions of species captured comparing the sampling times (χ2 = 35.7, df = 45, p = 0.83), showing that Ochotona pallasi pricei was always captured in highest numbers. Birds were captured in 0.98 % of the cases. Pikas were captured in 94.75 %, C. migratorius in 3.93 % and M. unguiculatus in 0.33 % of the cases. Pikas were thus always encountered 22 times more often than rodents. Out of 100 captured individuals the expected biomass of pikas would be 18.003 kg, of C. migratorius 0.142 kg, and of M. unguiculatus 0.017 kg. Pika biomass surpasses the sum of the other two small mammals by 113 times.

Discussion Mongolian pikas (Ochotona pallasi pricei) are by far the most abundant small mammals in terms of numbers as well as in terms of biomass. While pika numbers are about one order of magnitude higher than those of other small mammals, their biomass is two orders of magnitude higher. However, not using baits may have resulted in underestimating the abundance of rodents. On the other hand, rodents were captured even when using unbaited traps. Additionally, the frequencies of captured rodents were in accordance with observations during ﬁeld work. While pikas, pika droppings, and hay piles were apparent all over the place, there were few signs of other rodent activity, like droppings, runways or occasionally gathered seeds. These are concentrated in areas with a lot of gravel, stones and shrubs. Oenanthe isabellina also entered traps set on the ground. This bird is one of the most abundant species in the research area (own observation). They most probably use holes in the ground as nesting sites, which is supported by ﬁeld observation, since

31 2.4. GRID TRAPPING AND OBSERVATION 2. BASIC DATA COLLECTION

on some burrow entrances there were bird droppings. Even the chicks could be heard in one case. This is an example of a diﬀerent species using the structures provided by pikas (Smith and Foggin, 1999).

2.4.5 Welfare considerations in trapping and handling pikas Since this study is the first to have used life traps and individual marking on Ochotona pallasi pricei, it was not clear, if individuals would enter traps, how long they could stay inside a trap, and how they would respond to handling. It is not known, how long Ochotona pallasi pricei can stay within a trap without individuals being harmed. However, this information is crucial when planning capture sessions to set a maximum time span in which traps have to be checked. Even though they are not explicitly constructed as manipulative experiments, many field studies involve some degree of intervention: through disturbance caused merely by the presence of an observer, or where specific sampling techniques involve capture, handling, and marking. Close proximity to humans is necessarily traumatic for mammals. Putman (1995) lists three types of welfare costs for interventive fieldwork with mammals: (a) distress and mortality during capture, (b) distress and mortality during marking, (c) longer-term consequences of handling and marking in terms of subsequent (delayed) mortality or loss of fitness. The first two types of welfare costs are addressed in the following section.

Methods

Animals were captured from July 2000 to July 2002 using Sherman live traps (see Section 2.4.2, p. 28). Weight loss caused by trapping was calculated for two groups of animals: those captured twice within less than three days and those captured twice within three to 10 days. Mean weight loss of the two groups was compared to each other and to a 0 g weight change using a t-test. To assess, how long Ochotona pallasi pricei can stay in a trap without an individual being harmed, four animals were left in baited traps and released as soon as they showed weight loss. To assess further handling and trapping damage, it was noted when animals where found dead or bleeding in a trap, and their behaviour during handling was noted.

Results

Individuals lost more weight when captured within three days than when captured twice within more than three days. For 25 individuals a weight difference within less than three days could be calculated 37 times. Initial weights ranged from 144 g to 260 g. For 12 animals a weight difference within three to ten days could be calculated 44 times. Initial weights ranged from 52 g to 223 g. Weight change within less than three days ranged from 7.1 % to 6.9 %, with a median at 2.4 %. Weight change within three to less than 10 da−ys ranged from 1.7 % to 28.9 %, −with a median at 4.2 %. The maximum weight change of 28.9 % came−from the individual with the lowest initial weight, 52 g. Comparison of the two data sets showed that they come from different distributions (p < 0.001). Weight change within less than three days was significantly different from 0 % (p < 0.001).

32 2. BASIC DATA COLLECTION 2.4. GRID TRAPPING AND OBSERVATION

None of the four individuals observed in the traps accepted bait. They had to be released after 4.5 hours. Initial weights ranged from 147 g to 205 g. Weight change after 2.5 hours ranged from 1.36 % to 4.97 %, after 4.5 hours from 1.36 % to 5.3 %. − − − − Altogether, animals were captured and recaptured 362 times. Fifteen cases of animals bleeding at the front foot were observed and at least four of them had lost their claw within the traps. Three animals were found dead in the traps. One of them probably died because of heat stress, since it was a big individual in a small trap which was exposed to sunlight. This happened, although traps were checked every two to three hours during the day. At each handling session there were cases of squeaking or bleeding animals while their ears were marked. There were three cases of animals falling unconscious while being handled. Most of the animals urinated when they were held by humans. At the same time, most of the animals were quiet during handling and although few tried to bite, bites were not strong.

Discussion Sherman traps do not seem to be very suitable for Ochotona pallasi pricei. Being individuals of a burrowing species, they try to escape from the traps by digging themselves out, thus loosing claws and starting to bleed at the sharp metal edges of the traps. Wooden traps may offer a better choice. Since pikas were always interested in new objects, especially big objects, digging under them and sitting on them, it may be a good choice to use as big and heavy traps as possible (see next section on trap preference). It should also be kept in mind that pikas are very sensitive to heat stress. At least one of the individuals captured dead probably died of heat stress. It was a heavy animal captured in a smaller trap, so that it could not move inside the trap. Like all lagomorphs, pikas have high and constant body temperatures (around 40 ¢ ) in comparison to other small mammals. They are adapted to cold environments and cannot cope with high ambient temperatures (Matsumoto et al., 1995; Song and Fen, 1996; Wang and Wang, 1996). Knowingly sacrificing captured individuals, MacArthur and Wang (1973) found a lethal ambient temperature of 29 ¢ when exposed to it for more than 2 h. Traps should thus be protected from heating up. This is another reason to prefer wood over metal as material for traps. Not accepting food in captivity poses another difficulty when leaving traps open for a longer time. Since pikas were not active during nights, only the length of capture sessions during the day was crucial. Additionally, animals lost weight when they were recaptured on subsequent days. There should be a break of at least 10 days for the individuals to recover from a capture session. The weight gain of individuals captured within 2 to 10 days was probably due to still growing juveniles. Although most pikas were silent and did not move while being handled, three animal fell unconscious, and most of them urinated when being picked up. This probably represents signs of considerable stress for the animals. Other mammals, including hares, have been shown to be subject to trauma-related stress-shock, most commonly manifested in the form of a progressive post-traumatic myopathy in which rapid and irreversible changes in muscle tissues lead to prostration, progressive depression and death after a period of up to four days after capture (Putman, 1995 and references therein). One of the animals falling unconscious while being handled died within a short time. None of the other species died, although unbaited traps were left open during the night. Rodents captured were given food and they readily accepted it.

33 2.4. GRID TRAPPING AND OBSERVATION 2. BASIC DATA COLLECTION

Table 2.8: Individuals encountered during the capture and the observation sessions separately. See Table 2.6 on page 29 for more information on the sample sessions. Observation always included not identiﬁed individuals (not id.), where marks were absent or could not be assorted to a known individual. Numbers are given for female (f), male (m), and unsexed (u) individuals.

Capture Observation Nr. Capture date f m u Sum f m u not id. Sum 1 06.07.2000 10 2 3 15 2 17.07.2000 10 0 1 11 3 23.08.2000 6 2 0 8 4 28.09.2000 13 1 0 14 5 26.10.2000 12 4 0 16 6 28.11.2000 10 5 0 15 11 4 1 13 29 7 23.01.2001 6 1 0 7 12 3 2 14 31 8 23.03.2001 7 3 0 10 9 25.04.2001 9 2 0 11 12 4 1 14 31 10 22.05.2001 13 4 1 18 17 5 4 5 31 11 24.06.2001 34 13 15 62 19 6 11 7 43 12 27.07.2001 19 10 2 31 20 8 2 11 41 13 22.08.2001 14 4 0 18 18 9 0 6 33 14 23.09.2001 11 3 0 14 21 8 0 6 35 15 01.07.2002 19 3 0 22 15 4 0 5 24 16 07.07.2002 15 2 0 17 18 5 0 6 29

2.4.6 Capture and observation success This section ﬁrst lists the number of individuals encountered during all 16 capture sessions from July 2000 to July 2002 and 10 observation sessions from November 2000 to July 2002 (Table 2.6, p. 29). As this is the ﬁrst study to use life trapping on Ochotona pallasi pricei, some details on capture success regarding the size of the trap and the time of the day are given. Since the study used a robust design with several secondary sample sessions within a primary sample session, capture success in the secondary sample sessions could be compared. For 21 secondary sample sessions on 11 days a preference for trap size could be assessed. Traps used were Extra Large (10 12 38 cm) and Large × × (8 9 23 cm) Sherman live traps (Section 2.4.2, p. 28). Morning secondary sample sessions× × were usually longer than sampling sessions during the day, since traps were set in the evening before. For 16 morning secondary sample sessions (section 2.4.2 on page 29) capture success could be compared to the pooled data from the following secondary sample sessions on the same day. All of the pairs of morning and following day-sampling sessions were taken from the enlarged study site of (Table 2.6, p. 29 and Figure 2.11, p. 27). Preference for trap size and time of the day was analysed using a Wilcoxon signed rank test for paired comparison.

Results

Of the 155 individuals captured during the study period from July 2000 to July 2002, 94 could be assorted to female sex, while 39 could be assorted to male sex. For another

34 2. BASIC DATA COLLECTION 2.4. GRID TRAPPING AND OBSERVATION

22 animals sex could not be identified. Table 2.8 on the facing page lists the number of individuals captured and observed. Extra Large Sherman live traps were preferred over Large traps in the 21 secondary sample sessions analysed. Percentage of successful Large traps ranged from 0 % to 5.8 %, with a median at 1.5 %. Percentage of successful Extra Large traps ranged from 0 % to 14.8 % with a median at 3.9 %. Paired comparison of the percentages showed that the differences were significant (p = 0.005). Morning capture sessions were generally more successful than all the following capture sessions of that day together on all the 16 days analysed. Morning capture sessions had a median of 6.5 captures (5.4 % of the open traps), ranging from 2 to 27, while the sum of the day sessions ranged from 0 to 12, with a median at 5.5 captures (4.5 % of the open traps). Paired comparison over the 16 days showed that the differences were significant (p = 0.006).

Discussion Ochotona pallasi pricei proved to be a species easy to capture in comparison to other pikas. Puget (1971) tried to trap Ochotona rufescencs rufescencs without success. They captured animals by laying out nets over burrows and then digging out the whole burrow system. Kawamichi and Liu (1990) easily live-trapped O. curzoniae with meshed cate traps (13 10 33 cm) baited with apples and various ﬂowers, but × × they failed to capture O. cansus with traps. Individuals of this species had to be captured with a sweeping net usually used for collecting insects when leaving their burrow entrances. We repeatedly tried to capture Ochotona daurica with Sherman traps, with very little success. We could capture individuals inside our gers, when they had adopted them as home. In another occasion we dug holes in sand piles and thus holes in the ground. However, Ochotona pallasi pricei readily entered even unbaited traps. The results indicate that capture success might be enhanced by using larger traps: Pikas are very curious animals. They carried big, colourful things onto their burrows, collecting everything movable from ﬂowers and toilet paper to livestock dung. They liked to sit on, dig around and under traps or other objects new to them. Thus using larger, more prominent traps may enhance capture success. Trapping can possibly be restricted to the morning hours. The observed population entered traps more readily in the morning and by using this pattern, the critical period in the middle of the day, when traps may heat up, can be avoided. Additionally, captured individuals have time to recover.

2.4.7 To trap or not to trap? Comparing trapping with observation The following section assesses the eﬃciency of capture and observation in terms of encountering and identifying Mongolian pikas. For this, the accuracy of identifying marked individuals, the variability of encounters, and the information gained by pooling the data from both methods are analysed for the ten sample sessions, where data from both methods are available. Accuracy of individual identiﬁcation using ear tags and hair dye was measured by counting the recaptures being assigned to one known individual, more than one individual, or no known individual. Variability in individual encounters by observation or capture was compared using encounter histories for

35 2.4. GRID TRAPPING AND OBSERVATION 2. BASIC DATA COLLECTION 80

obs. only not id. obs. only id.

60 cap. and obs. cap. only 40 20 Number of individuals 0

1 2 3 4 5 6 7 8 9 10

Sampling occasion

Figure 2.12: Individuals of the Mongolian pika (Ochotona pallasi pricei) encountered only by capture (cap. only), only by observation (obs. only), or by both methods (cap. and obs.). Individuals encountered only by observation are again divided into those identified (obs. only id.) and those not having been identified (obs. only not id.). Dates for sample sessions were November 2000 (1), January (2) and April to September 2001 (3 to 8), and beginning and mid July (9 and 10) 2002. See Table 2.6 on page 29 for more information on the sample sessions. the 10 primary sampling sessions on which both capture and observation took place (Table 2.6 on page 29). The number of sample sessions between first and last encounters was compared for capture, observation, and pooling of the data. Regularity of encounters was measured by counting the number of sample sessions between two encounters. The percentage of encounters by each sampling method from the total number of encounters was calculated for animals having been encountered more than three times. To assess the additional information gained by either capturing or observation, the number of animals captured or observed were compared for each sampling session using linear regression. Additionally, relative encounter success was calculated for each method compared to the number of animals encountered by pooling data from capture and observation. Percentages were compared using a Wilcoxon test for paired data. A gain of information was quantified for each of the encounter methods. The information gained by observation are those individuals, which have been missed by capture during a sample session. Accordingly, the information gained by capture are those individuals missed by observation but captured. The number of individuals additionally encountered by observation or capture was evaluated using a χ2 test. Odds were calculated to compare the sample sessions (Quinn and Keough, 2002). Observed animals which were not identified were only taken into the analysis if they exceeded the number of animals captured, that were not observed.

Results The number of animals captured per primary sample session ranged from 7 to 62, with a median at 17.5. The minimum number of animals was captured in January 2001, the maximum in June 2001. The number of animals encountered by observation ranged from 24 to 43 pikas, with a median at 31. The fewest pikas were observed in June 2002,

36 2. BASIC DATA COLLECTION 2.4. GRID TRAPPING AND OBSERVATION

Table 2.9: Accuracy of identifica- Individuals Capture Observation tion of individually marked Mongo- assigned Rec. % Ind. Rec. % Ind. lian pikas Ochotona pallasi pricei by One 247 97.6 77 405 97.1 91 capture and observation. Recap- > 1 4 a 1.6 4 11 2.6 11 tured (Rec.) marked animals were None 2 a 0.8 2 1 0.2 1 assigned to one, more than one (> 1) or none of the previously encoun- Sum 253 100 83 417 100 96 tered marked individuals. Individu- a Animals did not have earplugs. als (Ind.) could be recaptured more than once. the most in June 2001. Pooling the data from capture and observation resulted in 26 to 72 encounters per primary sample session, with a median at 31 individuals. The minimum number of animals was encountered in June 2001, the maximum in June 2002 (Figure 2.12). Of the 155 pikas captured and marked with ear tags during the study, nine ear tags were seen on animals for more than a year. Colours of ear tags faded after a year. Fur dyed with picric acid stayed colourful for a few months in summer, and from October to March in winter. Probably due to their molt from summer to winter pelage, all fur markings were lost by the end of September, even if being newly marked. The combination of ear tags and patterns dyed in the fur resulted in an equally high accuracy of individual identification in capture and in observation. 83 animals were recaptured 253 times (Table 2.9). In 97.6 % of the cases, recaptured individuals were assigned to one individual only, which means that they were recognised without ambiguity. In 1.6 % of the cases there were two or more individuals with markings resembling the same picric pattern or ear cuttings of the animal recaptured. In 0.8 % of the cases it was not possible to assign a previously known individual. These two animals only had ear cuttings, no ear tags and no picric dyed hair. 96 animals were resighted 417 times during observation. With the binocular, ear cuttings and hair cuttings could not be used to distinguish individuals, nor could numbers on ear-tags be read. In 97.1 % of the cases resighted individuals were assigned to one individual each. In 2.6 % of the cases there was more than one individual with markings resembling the fur pattern and ear tag colours of an observed animal. In 0.2 % it was not possible to assign a previously known pika (Table 2.9). Individuals were more easily observed than captured. 144 individuals were encountered by capture or observation during the 10 sample sessions. 133 of them were captured, 99 were observed. There were animals marked before the first sampling session resulting in some individuals being observed only. The number of sample sessions between first and last captures ranges from one to five sample sessions. Observation and pooling of the data yielded a range of one to the maximal number of all 10 sample sessions. The distribution of the encounter time spans was strongly right skewed with a median of one for encounters by capture. More than half of the individuals were captured only once. Less than half of the individuals were encountered only once by observation, but the median was not much higher (Table 2.10). Most of the animals were observed and captured subsequently. The number of sample sessions between two encounters, which is the number of sample sessions an animal was missing between two encounters, ranged from 0 sample sessions, meaning subsequent captures (saturated encounter histories), to 4, 5, and 6 missing sample sessions

37 2.4. GRID TRAPPING AND OBSERVATION 2. BASIC DATA COLLECTION

Table 2.10: Length and regularity of encounters by capture, observation, and pooling both methods. Length of encounter period is the number of sessions between ﬁrst and last encounters. Regularity of encounters is the number of sessions between two encounters, in which this individual was neither captured nor observed.

Length of encounter period Regularity of encounter Method n Range a Median n Range b Median Capture 133 1-5 1 49 0-4 0 Observation 99 1-10 2 54 0-5 0 Pooled 144 1-10 1 70 0-6 0

a Maximum possible range is 10 sampling periods. b A value of 0 means no sampling periods between two encounters, thus subsequent encounters, while the maximum possible range is 9, for individuals marked before the first occasion and only encountered at the last (10th) occasion. for encounters by capture, observation, and the pooled data respectively. Distributions were strongly right skewed. All of the encounter methods had a median of 0 sample sessions between two encounters, meaning that more than half of the individuals were encountered subsequently by capture, observation, and pooling of the data (Table 2.10). 29 animals were encountered more than 3 times. They were captured 0 % to 100 % of their encounters, with a median at 50 %, while they were observed 40 % to 100 % of their encounters, with a median at 100 %. Paired comparison showed a significant difference of the distributions (p = 0.002). There was a greater chance to observe a Mongolian pika than to capture it. There was no simple linear relationship between the number of observed animals and the number of captured animals within one sample session (linear regression, r2 = 0.494), but a higher percentage of the individuals was encountered by observation than by capture (paired Wilcoxon signed rank, p = 0.037). The proportion of captured animals ranged from 20 % to 79 %, with a median at 48.5 %. The proportion of observed animals ranged from 54 % to 92 % with a median at 85.5 %. In all sample sessions both sampling methods, capture and observation, contributed additional information on the presence of individuals. The number of individuals not captured but additionally observed ranged from 9 to 29 animals, the number of individuals not observed but additionally captured ranged from 3 to 36. The proportions of individuals captured in contrast to being additionally observed differed for the 10 sample sessions (χ2, p < 0.001). The pattern of odds for individuals being captured instead of being additionally observed for the periods was complicated. The sample session of June 2001 differed most from the other sessions. There was a higher chance for individuals to be captured instead of being additionally observed in June 2001 than in all the other sampling sessions except for June 2002. January 2001 differed from all but 2 of the other sample sessions. There was a lower chance to be captured instead of being additionally observed in January 2001 than in most of the other sample sessions. In April 2001 and September 2001 the chance to be captured was not significantly higher than in January 2001 (Table 2.11 on the next page). The chances to be captured were 0.23 in January 2001, while the chances to be additionally observed were 0.77. In June 2001, the chances to be captured were 0.86, while the chances to be additionally observed were 0.14.

38 2. BASIC DATA COLLECTION 2.4. GRID TRAPPING AND OBSERVATION

Table 2.11: Information gain by comparing the chance to be captured with the chance to be additionally observed, but not capured. “+” indicates that chances to be captured were significantly greater than the chance to be additionally observed comparing the sample session of the row with each other sample session in the columns. “–” indicates that the chances were lower, while “ ” indicates no significant difference between the sample sessions. · Numbers in the column heading correspond to the numbers of the sample sessions in the row heading. The number of times with greater (odds > 1), smaller (odds < 1), and greater or smaller (sum) chances to be captured are summarised at the end of each row.

1 2 3 4 5 6 7 8 9 10 < 1 > 1 Sum 1 Nov 00 + – – 2 1 3 · · · · · · 2 Jan 01 – – – – – – – 7 0 7 · · 3 Apr 01 – – – 3 0 3 · · · · · · 4 May 01 + – 1 1 2 · · · · · · · 5 Jun 01 + + + + + + + + 0 8 8 · 6 Jul 01 + + – + 1 3 4 · · · · · 7 Aug 01 + – 1 1 2 · · · · · · · 8 Sep 01 – – – 3 0 3 · · · · · · 9 Jun 02 + + + + 0 4 4 · · · · · 10 Jul 02 + – 1 1 2 · · · · · · ·

The proportions of individuals observed in contrast to being additionally captured also diﬀered for the 10 sample sessions (χ2, p < 0.001). The pattern of odds was homogenous in this case. All of the sample sessions only diﬀered from June 2001. The chance to be observed instead of being additionally captured was lower in June 2001 than in all the other sampling sessions (Table 2.12 on the following page). The chances to be observed were 0.50 in June 2001. In the other months the mean chance to be observed was 0.82, while the chance to be additionally captured was 0.18.

Discussion According to Grunell and Flowerdew (1990), working with small mammals rarely includes direct observation. The Mongolian pika shows quite the contrary. The results from this study prove that data from direct observation is more reliable than capture data as far as the chance to encounter an individual is concerned. Using ear tags and picric acid to dye the hair of the pikas ensures an equally high accuracy of identiﬁcation as capturing, even though animals were observed up to distances of 50 m. The ear tags used in this study are not recommended for further use, though. See Kawamichi and Liu (1990) for ear tagging pikas. A gradual summer molt facilitates the use of dyed hair for identiﬁcation. In September, by the time of the winter molt, the change is more rapid and the dyed patterns get lost. Thus, it can be recommended to rely on observation data alone when using methods that involve re-encounters of marked individuals. Individuals were encountered more regularly (Table 2.10, p. 38) and, although capture added information, the number of individuals additionally captured stayed constant (Table 2.12, p. 40). The only exception was June 2001 where more animals were captured than in any other months. This was due to the many juveniles entering traps during that capture session. There were less juveniles in the summer of 2002, when overall abundance was also lower. This was

39 2.5. PARASITES 2. BASIC DATA COLLECTION

Table 2.12: Information gain by comparing the chance to be observed with the chance to be additionally captured, but not observed. “+” indicates that chances to be observed were significantly greater than the chance to be additionally captured comparing the sample session of the row with each other sample session in the columns. “–” indicates that the chances were lower, while “ ” indicates no significant difference between the sample sessions. · Numbers in the column heading correspond to the numbers of the sample sessions in the row heading. The number of times with greater (odds > 1), smaller (odds < 1), and greater or smaller (sum) chances to be observed are summarised at the end of each row.

1 2 3 4 5 6 7 8 9 10 < 1 > 1 Sum 1 Nov 00 + 0 1 1 · · · · · · · · 2 Jan 01 + 0 1 1 · · · · · · · · 3 Apr 01 + 0 1 1 · · · · · · · · 4 May 01 + 0 1 1 · · · · · · · · 5 Jun 01 – – – – – – – – – 9 0 9 6 Jul 01 + 0 1 1 · · · · · · · · 7 Aug 01 + 0 1 1 · · · · · · · · 8 Sep 01 + 0 1 1 · · · · · · · · 9 Jun 02 + 0 1 1 · · · · · · · · 10 Jul 02 + 0 1 1 · · · · · · · ·

probably caused by the drought summer of 2001, with one year’s time lag (Section 4.5, p. 101). Still, capturing pikas will be necessary for marking individuals. Another advantage of data based on observation is that even individuals avoiding traps can be recorded. A considerable part of the animals observed could not be identified (Figure 2.12, p. 36). These animals obviously avoided the traps, since they were never marked. Information on these animals would also have been missed, if trapping was used as the only encounter method. Thus, although combining sampling methods lead to some additional information, capture was less efficient than observation in order to gain information on individuals. Depending on the question asked, the sample size of an analysis can be enhanced by putting more effort into observation than into capture-recapture.

2.5 Parasites

Parasite load is related to survival inasmuch as it can be an indicator of individual health (Begon et al., 1996). In the following the number of pika individuals with parasites stand as proxy for population health during the year.

Methods Data was collected from animals captured within the 16 sampling periods (Table 2.6, p. 29) between July 2000 and July 2002. Parasites were noted when they were evident without explicitly searching for them. Fleas, fly larvae, and ticks were counted, while mite data was recorded as presence/absence data. While fleas, ticks, and mites are ectoparasites, the larvae of the fly were always found within the limbs of the Mongolian pikas. They probably belong to the species Oestomyia leporina (Pallas, 1778; Mináˇr and H˚urka, 1980).

40 2. BASIC DATA COLLECTION 2.5. PARASITES

2000 15 5 0

2001 15 5 0

Number of individuals 2002 Flea 15 Fly Mite

5 Tick 0

Jan Mar Apr May Jun Jul Aug Sep Oct Nov

Month of capture occasion

Figure 2.13: Number of pika individuals with parasites during the sampling periods from July 2000 to July 2002. Months with capture occasions are labelled with “ ”. ◦

Results Flea numbers per individual ranged between 1 and 29, with a median at 2. Numbers of fly larvae per individual ranged between 1 and 4, with a median at 1. Tick numbers per individual ranged between 1 and 6, with a median at 1. The highest numbers of fleas per individual where observed when days got shorter in winter and handling of animals continued after sunset. Fleas then jumped from the animals, while during the day they were not as easy to detect. After this happened, handling animals after sunset was stopped. Fleas were present throughout the whole year (Figure 2.13). In January 2001 however, no parasites were detected. Flies were present from June to November, although the number of animals with flies in November 2000 was low. Mites were detected in the months of June to September in all three years, ticks in July and August. In winter, parasite presence on the animals was thus lower than in summer.

Discussion Although in winter conditions for living are adverse in the Gobi desert, since temperatures are low and there is no fresh fodder for the animals, pikas cope well with these conditions. The highest number of individuals with parasites were reached in the summer months, indicating poorer health in summer. This is in accordance with the results on survival rates presented later in this study. As will be shown there, survival of pikas decreases with density, which is high during reproduction in summer, while survival in winter is high (Section 4.2, p. 79).

41 2.6. MOVEMENT 2. BASIC DATA COLLECTION

2.6 Movement

In connection with the concern that the pikas occurring in the National Conservation Park may be small mammal pests it was asked, how fast populations could spread. As a ﬁrst approximation to answer this questions, movement distances of individuals were taken.

Methods For this, not only the data from the capture and observation from all sample sessions (Table 2.6, p. 29) were used, but also any information gathered outside these sample sessions on the trapping site. Additionally observation took place in a 25 m belt around the study area in September 2001 and July 2002. The maximum distance which could have been observed was thus the diagonal, 212 m. For each encountered individual, the maximum distance moved during a specific time interval was calculated. Three different time scales were analysed for male and female individuals separately: the entire sample period from June 2000 to July 2002, the time intervals between subsequent sample sessions and the movement within one sample session. To avoid a bias towards short movement distances caused by capturing individuals only a few times, only individuals having been encountered at least five times have been taken into the analysis at the highest time scale. To explore seasonal- ity and shorter time intervals, within and between sampling movement distances were calculated. Time intervals within sampling sessions ranged between one and 10 days, while time intervals between sampling sessions ranged between 10 days and 10 months, although most of the between-sampling sessions time intervals were one month (Ta- ble 2.6, p. 29). Animals captured or observed at least twice within one sampling session were used for analysis. Between-sampling-session movement was calculated from animals observed or captured at least twice between one sampling session and the previous one. Eighty three animals could be captured or observed at least five times, of which 60 were female and 23 male. For half of the animals, the time interval observed was less than 86 days, the longest time observed for one animal was 738 days, more than two years, and the shortest time was one day. Females were observed within time intervals from 1 to 738 days (median 102 days), males from 1 to 352 days (median 62 days). The number of female and male pikas captured within sampling sessions ranged between 4 and 36 for females, and between 1 and 10 for males. The numbers between sampling sessions ranged between 4 and 22 for females and 0 and 9 for males.

Results Distances moved within the time interval of two years ranged within 4.2 and 96.8 m. Median of the maximum distances moved was 30 m. Females moved within 4.2 and 96.8 m, the median was 31.5 m, males moved within 5.4 and 81.2 m, the median was 22.2 m (Figure 2.14). The female which had moved 96.8 m had been observed within 146 days. Within sampling sessions, maximum distances moved were generally less than 20 m for females and males (Figure 2.15). Only in April 2001 and June and July 2002 males moved more than 20 m. Between-session movement was consistently higher than within-session movement for both sexes, but the diﬀerence was not large. Between- session movement was generally between 20 and 30 m for females and males. Exceptions were the months of March to May in 2001. In these months, males moved more than

42 2. BASIC DATA COLLECTION 2.6. MOVEMENT

Females Males 15 15 10 10 Frequency Frequency 5 5 Frequency 0 0

0 20 40 60 80 100 0 20 40 60 80 100

Maximal distanceMaximal moved distance moved

Figure 2.14: Number of individuals with comparable movement distances (in meters) within a time interval of two years, from July 2000 to July 2002. Only individuals encountered at least 5 times were included. The maximum potentially observable distance was 226 m, but no animal moved that far.

30 m as a median moving distance. Between April and May all animals observed moved a maximum distance exceeding 40 m. The single male observed between March and April also moved more than 40 m.

Discussion The results indicate that populations of Ochotona pallasi pricei can be expected to show rather low mobility. Movement distances of individuals increased when looking at time intervals of some days in contrast to some months, but they did not increase when extending the time interval to two years (Figures 2.14 and 2.15). The movement distances observed corresponded well with the nearest neighbour distances between two burrows. While the average nearest neighbour distance was 14 m (Section 2.4.1, p. 26), the average movement distance was 30 m. These results are in accordance with the movement distances reported for other pika species. In a review on pika dispersal patterns, Smith (1987) defined movement distances of 50 m already as dispersal, since the observed species did not move far. Observational data on the non-burrowing species Ochotona princeps suggest that juvenile settlement patterns are primarily philopatric; adults on neighbouring territories form monogamous mated pairs, and pikas disappearing from a population generally die rather than disperse successfully (Smith, 1987). These conclusions suggest that pikas inbreed regularly with close relatives. Hafner (1994) stated that for O. princeps recolonisation within 20 km of pika populations was rare. Nevertheless single individuals move further. Peacock and Ray (2001) report dispersal distances up to 396 m for Ochotona princeps. Despite philopatric settlement pattern, no clusters of highly related individuals could be found using genetic fin- gerprinting (Peacock, 1997). Based on observational data they showed that juveniles maturing early in the season settle philopatrically, while those maturing later travel further to find a vacant territory. O. princeps thus showed a dispersal pattern congruent with the dispersal model of competition for resources with territory as resource.

43 2.6. MOVEMENT 2. BASIC DATA COLLECTION

01−23 80 Female

40 Male 0 03−23 80 40 0 04−25 80 40 0 05−26 80 40 0 07−06 06−28 07−02 80 40 0 07−17 07−30 07−11 80 40 0 08−23 08−25 80 Maximal distance moved [m] 40 0 09−28 09−25 80 40 0 10−26 80 40 0 11−28 80 40 0 within between within between within between 2000 2001 2002

Figure 2.15: Movement of female and male individuals within sample sessions (a couple of days) and between a given sample session and the preceeding one (generally one month) for sample sessions from June 2000 to July 2002. The three diﬀerent years are presented in columns, while rows indicate similar dates within each year. Sampling dates are given in the upper left corner of each plot. Male and female movement diﬀered from March to May 2001, when males moved longer distances between sample sessions.

44 2. BASIC DATA COLLECTION 2.6. MOVEMENT

Seasonal differences in movement pattern and territory size have been reported for several pika species. For Ochotona daurica and O. curzoniae regrouping of territories was observed from April to March (Wang et al., 2000; Wang and Dai, 1990). Kawamichi (1971) noted higher activities for males in spring in Ochotona hyperborea yesoensis. He differentiated four phases in the yearly cycle of pika behaviour: a spring phase with higher activities of males, then a family phase in summer which again is followed by a regrouping phase and finally a stable phase in winter. In a study on movement pattern of Ochotona alpina, Nikol’skii and Mukhamediev (1997) showed that home ranges increased during summer, as well as the overlap between home ranges. Males generally had larger ranges, but the difference got smaller at the end of summer. Range overlap was higher in summer than at the end of summer. Home ranges varied between 404 and 5783 m2. This corresponds to circles with radii of 11 to 43 m. For the Mongolian pika (Ochotona pallasi pricei) Okunev and Zonov (1980, quoted in Smith et al., 1990) note that mean territory size decreases from summer to winter and with increasing density. Territory size was 403 m2 in August. This corresponds to a circle with a diameter of 23 m. This is a distance surpassing the median distances moved by individuals of this study, but is well in accordance with the between-sampling- sessions movements (Figure 2.15, p. 44). Territory size is up to 129 m2 in January, depending on snow depth (Okunev and Zonov, 1980), which would correspond to a circle with a diameter of 13 m. The present study observed higher male mobility in spring and higher mobility in the summer of 2002, when density was lower (Section 3, p. 47). Higher male mobility in spring indicates a polygamous mating system.

45 2.7. SUMMARY 2. BASIC DATA COLLECTION

2.7 Summary

The mountain steppes of the Gobi Gurvan Saikhan National Conservation Park (GGS) in the Mongolian Gobi Altai host rare and endangered animal species like the snow leopard (Uncia uncia), and the argali (Ovis ammon). Additionally they are an important rangeland for livestock of local herder families. The Mongolian pika (Ochotona pallasi pricei), a subspecies of the Pallas pika, is the dominant small mammal species in the mountain steppes of the GGS. It outnumbered other small mammals by one and outweighed them by two orders of magnitude. The life history patterns of the genus Ochotona were reviewed showing that this species could be expected to reach high population densities and may exhibit traits of a small mammal pest. This study presents the first account on trapping and observing individually marked Mongolian pikas. It therefore presents instructions on how to capture, handle, and observe individuals of this species. Mongolian pikas can be captured in conspicuous and preferably large life traps without bait, traps should be checked every two to three hours during the day to prevent heat stress but can be left open during the night since the pikas do not enter traps in the dark. The morning hours were the most efficient time of the day for capturing. Using ear tags and picric acid as hair dye the animals could be observed and identified up to a distance of about 50 m with binoculars. It is recommended to use an elevated look-out protected from wind. Data based on observation was more reliable in terms of encounter success than data based on capture. A map of a mountain steppe area with burrows and observation on burrowing activities shows that pika burrows are large structures probably lasting for centuries. A reduced parasite load during winter suggests that this season, despite of extreme climate conditions, does not present a challenge for pika health. Usual movement distances of pika individuals corresponded to the distance from one burrow to the next. Populations of this species are thus not expected to spread fast.

46 Chapter 3

Estimating Density

The impact of a species on the ecosystem is related to the number of individuals active within an area, thus, to its population density (Begon et al., 1996; Putman, 1989). At the same time, survival rates are often influenced by densities (Begon et al., 1996; Krebs, 1985; Smith, 1988). This chapter therefore estimates the density of the observed population of the Mongolian pika (Ochotona pallasi pricei). Commonly used methods for density estimation include the estimation of abundance and the estimation of the effective area from which samples were taken. This chapter first uses the number of pika individuals actually observed to assess the usefulness of commonly employed abundance estimators for the Mongolian pika. It then uses movement distances to deduce an effective area from which samples were taken to finally estimate pika densities within the sampling period.

3.1 Introduction

A large number of models has been developed to estimate abundance from repeated sampling occasions of individually marked animals. Frequently used techniques include the Peterson (1896), Schnabel (1938), Schumacher-Eschmeyer (1943), and Jolly-Seber (Jolly, 1965) estimators (Krebs, 1999). All these models calculate their estimates based on the assumption that

1. the probability of an individual to be sampled is the same for all individuals during one sample session and

2. that there are no losses or additions of individuals between sample sessions.

Thus, the proportion of individuals caught from the marked population in a second sample (M2) to the number of individuals captured in the second sample (N2) resembles the proportion of the number of animals captured and marked in the ﬁrst sample (N1) to the total number of individuals present (N).

N M 1 = 2 (3.1) N N2 Closed population estimation models assume that there is no loss or addition of animals during the time of the study. Common estimation models include the Petersen method for two successive sample sessions (Petersen, 1896) and the Schnabel method

47 3.2. METHODS 3. ESTIMATING DENSITY

(Schnabel, 1938) for more than two successive sample sessions. The Schumacher- Eschmeyer method (Schumacher and Eschmeyer, 1943) is a derivation of the Schnabel method, using linear regression analysis to calculate estimates. The Jolly-Seber method is an open population estimation model (Jolly, 1965). It allows for loss and addition of animals during the time of the study and uses the information on the animals’ appearance in sample sessions before and after a given session to estimate the number of marked animals present. Although the assumptions of equal catchability and no loss or additions are unre- alistic in most cases (Krebs, 1999) and have been shown to underestimate abundance (Carothers, 1973; Koper and Brooks, 1998; Manning et al., 1995), they continue to play an important role in scientiﬁc and management practise. Researchers simply state that although assumptions were probably violated, bias was minimised wherever possible (Caughley, 1977; Koper and Brooks, 1998). All studies on pikas known to the author did not rely on any of these estimators, but simply stated either that all animals of a given population have been captured and marked (for example Wang and Smith, 1988), or relied on counting unmarked individuals (for example Leont’ev, 1968; Zevegmid, 1975). Because of their diurnal activity pattern and the open steppe environment Ochotona pallasi pricei are conducive to complete enumeration by observation for assessing abundance within a well deﬁned area. In the following the abundance estimators introduced above are tested on the observed population of the Mongolian pika and compared to the number of individuals actually observed.

3.2 Methods

To compare the density estimates, data from the 10 sample sessions with combined trapping and observation from November 2000 to July 2002 were used (Table 2.6, p. 29). The size of the study site was 100 100 m2 (Figure 2.11, p. 27). Sampling eﬀort was higher for capture than for observation.× While two days were necessary for capture recapture, observation was usually done in about two hours.

Abundance The estimators of Schnabel and Schumacher-Eschmeyer deal with robust sampling designs, where primary sampling sessions include several secondary sample sessions. These were calculated for capture data alone, since observation only consisted of one sampling session. Likewise the Jolly-Seber technique could only be used for capture data, since there were a considerable amount of not identified individuals in the observation data set, which could not be considered as newly marked individuals as requested by this method. The Petersen estimator could be calculated for both data sets, taking the capture session as the first and the observation session as the second sample. These estimates were compared to the minimum number of individuals alive (MNA) encountered by capture or observation. For this, capture and observation data were combined by adding not observed but additionally captured individuals when they exceeded the number of not identified individuals (table 2.8 on page 34). Software provided by Krebs (1999) was used for the calculations.

Eﬀective area Eﬀective sampling areas (Aˆ) have to be estimated separately for grid- based mark-recapture data to deduce density (Dˆ) as number of animals (Nˆ) per area

48 3. ESTIMATING DENSITY 3.3. RESULTS

(Krebs, 1985; Parmenter et al., 2003). Estimated eﬀective area (Aˆ) was derived from the equation Aˆ = L2 + 4LWˆ + πWˆ 2, (3.2)

where L is the length of one grid side (100 m) and W is a boundary strip related to the movement pattern of the animals (Krebs, 1985). There are several approaches to estimate sizes of boundary strips (Wˆ ) to account for the effective sampling area including inter-trap distances, the animals’ home-range sizes, or average animal movements among traps (Krebs, 1985; Parmenter et al., 2003; Wilson and Anderson, 1985). Although using average movements among traps is hampered by relying on the small sample sizes (Mares et al., 1980) and the nonrandom selection of animals entering a trap more than once within a short period of time, they have been shown to perform well empirically. Parmenter et al. (2003) used full mean maximum distances moved among traps for their best estimate for effective sampling area. The width of the boundary strip in the present study was thus estimated using maximum movement distances of all re-encounters detected by capture, observation, and pooled data within the primary sample periods. In the pooled data the maximum movement distance could result from two encounters through capture, two encounters from observation, or from capture and observation of an individual. It was assumed that each individual had moved at least one meter before reaching the trap. Maximum distances moved of 0 m were thus replaced with 1 m. If the maximum distance moved for a given encounter method was smaller than half of the inter-trap distance, this distance was used instead. Half of the inter-trap distance was 5 m in the present study, resulting in a minimum effective area of 1.2079 ha. The influence of sampling method and sampling period on maximum movement distances of re-encounters was evaluated using an analysis of variance with two variables, three factor levels of methods and ten factor levels of sample periods. A fourth-root transformation was used to achieve normality in the residuals. The following models were compared: The full model, the model without interaction, and the reduced model with the method factor levels reduced to capture as one level and both observation and pooled data as the second level. The resulting model was compared to three different groupings of the time factor levels, a) winter (November 2000 and January 2001) against non winter, b) winter against non winter 2001 and non winter 2002, and c) the three years separately. Models were compared using likelihood ratio tests. The medians of the resulting subgroups of the most parsimonious model were used as boundary strip width.

Density Density (Dˆ) was then estimated based on the abundance estimates (Nˆ; compare Equation 3.1) and the eﬀective area (Aˆ; Equation 3.2) according to the equation

Nˆ Dˆ = , (3.3) Aˆ

where eﬀective area depended on the maximum distance moved for a given sample period and sampling method.

49 3.3. RESULTS 3. ESTIMATING DENSITY P T a capture etersen able Jan No Ma Apr Jun Jul Jun Sep Aug Jul lo w v y er 02 01 01 01 01 02 01 00 01 01 3.1: (lo) data, estimators. 15 18 11 62 17 22 14 18 31 and Abundance 7 c while 29 31 31 43 29 24 35 33 41 31 upp o MNA the er 29 31 31 72 29 35 26 33 41 31 MNA p (up) P (Ab) etersen 20.6 19.8 19.8 46.1 16.3 22.4 14.6 21.1 26.2 22.0 abundance D conﬁdence and 117.6 estimator 64.7 16.0 22.9 36.0 14.5 densit D - - - - estimates in 161.8 Jolly-Seb y 89.1 31.5 49.6 17.5 terv Ab (D) 22 is - - - - al based estimates 96.3 42.1 23.1 10.3 15.2 (95 are 39 er lo %) on giv 408.3 316.7 data can 58.6 48.1 45.2 79.6 en up for for b from e minimal 61.0 22.3 22.1 14.1 12.4 35.5 20.2 13.8 19.6 18.7 giv captured D p en. o oling 83.9 30.7 30.4 48.8 24.4 28.3 25.8 Ab Jolly-Seb Sc n 17 17 20 um hnab (c), capture b 58.2 14.4 15.5 29.8 10.9 17.2 10.7 11.3 er 3.2 8.6 observ el lo aliv er, 125.1 333.3 and e Sc 70.4 64.8 84.3 36.2 61.9 75.4 37.6 ed (MNA), up 55 hnab (o), observ 60.5 24.4 23.7 34.5 13.4 11.6 23.4 20.1 13.9 20.3 el, and D ation. and Jolly-Seb p 83.2 33.6 32.6 47.5 16.2 15.9 28.3 27.6 20.1 29.4 o Sc Ab Sc oled h h u-Esc umac er, 63.6 19.8 19.9 34.8 10.1 14.6 20.2 (p) 1.9 9.7 9.6 lo Sc h her-Esc data. hnab 120.1 113.1 244.8 89.9 75.2 45.2 54.3 up el, F hmey - - - or Sc 65.0 27.1 28.8 40.3 26.7 59.9 28.1 19.7 33.4 22.8 all h er D umac estimators estimates 101.7 42.4 45.1 41.7 84.3 34.9 40.5 her-Esc P Ab 63 44 47 etersen 86.3 34.3 36.3 52.7 30.8 32.3 29.7 34.3 32.4 except are hmey 35 lo based er, 129.2 187.9 59.8 79.7 61.3 44.6 58.2 72.7 54.8 MNA up 56 and on

50 3. ESTIMATING DENSITY 3.3. RESULTS

3.3 Results

Abundance The minimum number alive of pikas encountered on the trapping site (MNA) varied between 26 and 72 animals (difference = 46, Table 3.1). Highest abundance was reached in June 2001 and lowest in June 2002. Abundance estimated by the Peterson technique for the data pooling observation and capture sessions yielded higher numbers than the MNA (Table 3.1). Numbers ranged between 34.9 and 101.7 individuals, the difference amounting to 76.8 animals. Highest abundance was reached in June 2001 and lowest in June 2002. The MNA were lower than the 95 % confidence intervals in eight of the ten sample sessions, exceptions were April 2001 and Septem- ber 2001. The confidence interval for January 2001 (ranging from 32.3 to 187.9 individuals, a difference of 155.6 individuals) was larger than the remaining intervals with differences between 14.9 and 42.9 animals. Width of confidence intervals ranged from 42 % of the estimated abundance in June 2001 to 185 % of the estimated abundance in January 2001. The median was 52 % of estimated abundance. The following estimates were based on capture data alone. Variability of estimates was highest using the Jolly-Seber estimator (Table 3.1). Abundance varied between 17.5 and 161.8 animals (difference = 144.3). Highest values were reached in May and June 2001 (89.1, 161.8 individuals) and lowest values for January and April 2001 (17.5, 22.0). Except for the May and June estimates in 2001 MNA was within the 95 % confidence intervals. Estimates for May and June were higher than MNA. Confidence interval widths for May and June (274.6, 312) were higher than in the remaining sampling periods (30.0–40.6). Width of confidence intervals ranged from 82 % of the estimated abundance in July 2001 to 308 % of the estimated abundance in May 2001. The median was 165 % of estimated abundance. Schnabel and Schumacher-Eschmeyer estimates yielded very similar results (Table 3.1). Variation was higher than MNA. Abundance was estimated to have been high in June and July 2001, and low in the January and April 2001 and July 2002. All estimates, except for June 2001, were lower than the MNA. The 95 % confidence intervals of both methods always included the MNA, but high confidence interval widths showed low precision of the estimates. For the Schumacher-Eschmeyer estimate, the upper confidence limit exceeded the possible range in three cases, in November 2000, January 2001, and June 2002. The remaining widths of confidence intervals ranged between 34.1 and 230.2 animals. The widths of the 95 % confidence intervals of the Schnabel method varied similar to the Peterson estimator: A high value for January 2001 (330.1 individuals), and lower values for the remaining sampling periods (26.3–66.9 individuals). Width of confidence intervals in the Schnabel estimates ranged from 80 % of the estimated abundance in June 2001 to 1942 % of the estimated abundance in January 2001. The median was 162 % of estimated abundance. The smallest width of confidence intervals in the Schumacher- Eschmeyer method was 68 % of the estimated abundance in June 2001. Counting the three sample sessions, where the upper limit could not be calculated, as infinitively high estimates, the median was 251 % of estimated abundance.

Movement distances. Using capture data, 65 maximum movement distances could be calculated. The number of distances within the sample periods ranged from 1 in January 2000 to 16 in June 2001. Observation yielded 108 maximum distances, ranging from 3 in August 2001 to 22 in September 2001. From the pooled data 182 maximum distances could be calculated, ranging from 10 in January 2001 to 26 in September 2001.

51 3.4. DISCUSSION 3. ESTIMATING DENSITY

Table 3.2: Movement distances and eﬀective sampling area for capture and observation.

Captured Observed or Pooled Distance [m] Area [ha] Distance [m] Area [ha] Winter 1.0 1.2079 9.5 1.4084 Summer 01 8.8 1.3763 12.8 1.5635 Summer 02 10.3 1.4453 17.1 1.7759

The analysis of variance showed that both, the sampling method (p < 0.001)and the sample period (p < 0.001), had a significant effect on maximum movement distances, while the interaction was not significant. Removing the interaction from the model resulted in an equally good description of the data (p = 0.99). In a next step the model without interaction was reduced to two levels of sampling method: capture vs. observation with pooled data. This model could explain the data equally well (p = 0.75). The model could be further simplified by grouping the time factor as described above according to model b). This model differentiated three groups, namely Novem- ber 2000 and January 2001 in one group, the remaining sample sessions in 2001 in the second group, and the sample sessions of 2002 in the third group. The other two models could not explain the data as well [model a): p = 0.02; model c): p = 0.07]. The resulting 6 subgroups of the most parsimonious model produced the following medians for maximum movement distances. Movement distances were lowest for captured animals in winter 2000/2001 with a median of 1 m. They were highest in summer 2002 for observed animals or pooled data with 17.1 m. For capturing, the movement distances increased from winter 2000/2001, to the vegetation period of 2001 with 8.8 m, to summer 2002 with 10.3 m. For observation and pooled data they also increased from winter 2000/2001 with 9.5 m, to the vegetation period of 2001 with 12.8 m, to the summer of 2002 (Table 3.2). The resulting effective sample areas associated with the movement distances are listed in Table 3.2.

Densities Accounting for movement distances reduced the differences between the sampling methods and enlarged the differences between the years 2000/2001 and the year 2002 compared to abundance estimates (Table 3.1, p. 50 and Figure 3.1, p. 53). Median density based on the MNA was 20.85 individuals per hectare, while first and third quartile were 19.8 and 22.4 individuals per hectare (Tukey’s five number summary, R Development Core Team (2004)). Densities were stable for most of the observed period of time, their interquartile range spanned 2.6 animals. While the density in June and July 2001 was above the interquartile range, the density in June and July 2002 was below this range.

3.4 Discussion

All of the employed abundance estimators produce reasonable estimates. The high conﬁdence intervals reﬂect the low sample sizes for the estimators based on capture data alone (Krebs, 1999; Manning et al., 1995; Seber and Schwarz, 1999). The number of animals captured during a primary sample session reached half of the MNA estimates

52 3. ESTIMATING DENSITY 3.4. DISCUSSION

120 MNA Petersen Jolly−Seber 100 Schnabel 80 60 Number Individuals 40 20 0

Jan Mar May Jul Sep

Date (in the years 2000 and 2001)

Figure 3.1: Estimated densities for the observed population of the Mongolian pika (Ochotona pallasi pricei) on one hectare. Minimum number alive (MNA) and Peterson estimates were based on pooling data from capture and observation, while Schnabel and Jolly-Seber estimates were based on capture data alone. For measures of precision see Ta- ble 3.1, p. 50. in some months, thus capture probability in the secondary sample sessions was well below the recommended probability of 0.5 (Hilborn et al., 1976). ≥ Although the Petersen estimator used the highest sample sizes and consequently produced the lowest confidence intervals, they don‘t contain the MNA. The estimates are still reasonable, since the MNA is a minimum number, it is possible that animals have been missed. But the estimates produced by the Petersen estimator are probably too high. They are based on the capture data, which has variable and comparatively low success, and the observation data, which has a high encounter success. The method is sensitive to the violation of the assumptions, since the especially low capture success in January 2001 is reflected in unrealistically high estimates for abundance. But at the same time, the confidence interval was increased, so that MNA was still near the lower limit. It is recommended to combine sampling methods in small mammal studies to enlarge sample size and to reduce variability in individual capture probabilities (Koper and Brooks, 1998; Krebs, 1999; White et al., 1982). However, in the case of the Mongolian pika the Petersen estimator might perform better when based on two successive observation sessions. In this case, only the marked population can be estimated, but combining two similar encounter methods will increase precision. Although the estimators based on capture underestimate abundance, they still perform much better than capture data alone. Even the high abundance estimated by the Jolly-Seber method reflect the reproductive period. This is the time, where observation

53 3.4. DISCUSSION 3. ESTIMATING DENSITY

misses many individuals, as has been shown in Section 2.4.7 on page 35 (Table 2.12, p. 40). This is consistent with other studies that show that estimation models based on capture-recapture perform better than relying on counting individuals Thus, although estimation models based on capture-recapture give reasonable results for abundance, this is hampered by the variable capture probabilities. It is better to rely on observation, counting individually marked animals. Using two observation sessions in succession overcomes the disadvantage of enumeration techniques, which cannot give measures of precision. Sampling effort is not increased much, since the sampling effort to observe is much lower than the effort to capture and mark the individuals. Median density based on MNA was 20.85 pikas per hectare, higher densities were reached in summer 2001 and lower densities in summer 2002. Since movement distances were greater in 2002, this again lowered densities in comparison to abundances. The strong contrast between densities in the summer of 2001 and the summer of 2002 reflects the drought conditions in the year 2001 (Section 2.1, p. 13, Retzer (2004)). While numbers are still high in the drought summer, they decrease in the year after the drought. The greater movement distances can be a reaction on the lower abundance in the summer of 2002. Mongolian pikas have been reported to show highly aggressive behaviour towards each other (Monkhzul, 2005; Proskurina et al., 1985). Greater movement distances with lower abundance indicate that movement distances are controlled by the territorial behaviour of the individuals. Median density was 20.85 animals per hectare, while the study site comprised 28 burrows (Section 2.4.1, p. 26 and Figure 2.11, p. 27). This is in accordance with the observation that especially very small “burrows” belonged to the territories of individuals mainly occupying adjacent larger burrows. Examples for such small burrows were those with the coordinates (18,8), (52,8), (53,15), and (85,64) in Figure 2.11 on page 27. Generally, one of the larger burrows is inhabited by one pika. High densities in summer are misleading when assessing the impact of pikas on the ecosystem, since the number of animals does not reflect the increase in biomass (Owen-Smith, 2005). Because of the many juveniles, mean body weight was lower in summer than in the other months (Figure 4.1, p. 60). The combined biomass of the Mongolian pikas is thus less different than numbers imply.

54 3. ESTIMATING DENSITY 3.5. SUMMARY

3.5 Summary

In the study period pika densities varied between 14.6 and 49.8 individuals per ha. Densities were stable for most of the observed period of time, their interquartile range spanned 5.7 animals. Median density was 21.4 animals, which is less than the 28 burrows on the trapping site. Density was higher in the drought summer of 2001, while it was lower in the summer after the drought, indicating a time lag in the response to the drought conditions. Movement distances were higher in the summer after the drought when density was lower. All of the four methods for estimating abundance produced reasonable estimates. Methods used were Schnabel, Schumacher-Eschmeyer, and Jolly-Seber for capture data and the Petersen estimator for data from capture and observation. The methods using capture data alone underestimated abundance, while the Petersen estimator, using both capture and observation, overestimated abundance. For further studies it is recommended to use successive observation sessions and then apply the Petersen estimator on observation data only to estimate abundance.

55 3.5. SUMMARY 3. ESTIMATING DENSITY

56 Chapter 4

Life history

Considering a small mammal to be a pest species is coupled with the fear that the animals might reach high population densities. Only one of the two life history strategies introduced in Section 2.2 on page 23 could possibly be considered to reach high population densities. Life history traits associated to this strategy include life span, age structure (Section 4.1), survival rates of juveniles and adults, especially the influence of density and climate on survival rates (Section 4.2, p. 79), and the reproduction (Sec- tion 4.3, p. 96) of the species. The chapter ends with summarising the information on survival and reproduction in a model on population dynamics for the Mongolian pika (Section 4.4, p. 99) and finally discusses the findings on survival and density dynamics for the life history pattern of Ochotona pallasi pricei (Section 4.5, p. 101).

4.1 Age structure and life tables

Age structure of a population is an important demographic parameter. When comparing the K-type life history pattern of non-burrowing pikas with the r-type life history pattern of burrowing species, non-burrowing species had lower juvenile and yearling ratios, due to high juvenile mortality (Smith, 1988, and Section 2.2, p. 23). Life tables first were developed by human demographers, particularly those working for life insurance companies, which have a vested interest in knowing how long people can be expected to live (Krebs, 1985). They are commonly used in ecological research to summarise and compare patterns of survival for all groups of organisms. Life tables are based on cohorts, which consist of individuals originating from the same period in time. They are thus dependent on the knowledge of how old an organism is. However, determining age of mammals is a difficult task. One approach to assess age is to monitor individuals from birth, observe juvenile growth and then decide upon a probable adult age. This can then be used to classify newly captured individuals and assort them to a probable year of birth. In the following section (Section 4.1.1, p. 58), this is done using two different criteria: a static and a dynamic weight limit. Having assorted individuals to cohorts makes it possible to study population age structure during the year (Section 4.1.2, p. 69). Finally, life tables can be used to summarise the fate of cohorts through the years (Section 4.1.3, p. 74).

57 4.1. AGE STRUCTURE AND LIFE TABLES 4. LIFE HISTORY

4.1.1 From weight to age Characteristics of the cranial bones and sutures have been used to age pikas, but for this animals have to be sacrificed (Krylova, 1973; Millar and Zwickel, 1972). A less intrusive approach is trying to identify a minimum adult weight and classify newly captured individuals according to this weight. Two different criteria to determine pika adult weight have been employed in this study, (1) a static criterion using minimum winter weight of all captured animals, henceforth the winter-weight criterion, and (2) a dynamic criterion, using juvenile growth curves, henceforth the growth-curve criterion. The latter is a more liberal approach, combining data on juvenile growth rates with data on weight structure for the capture sessions and observations from field work. These two criteria lead to two different results concerning which animals should be considered juvenile in which year. In the following the advantages and disadvantages of the two criteria are described. Using the winter-weight criterion there is a chance that juveniles may be regarded as yearlings. Minimum winter weight may be considerably lower than an average adult weight during the whole year, depending on how much weight is reduced during the winter months. This criterion thus theoretically underestimates the number of juveniles of a given year. Individuals reaching minimal winter weight already within their summer of birth will be considered yearlings at first capture, increasing the cohort strength of the last year, but not that of the given year. This scenario is plausible, since individuals probably loose weight again in winter. On the other hand, minimum winter weight is a criterion easy to use and to reproduce. Additionally, pikas might be expected to loose little weight in winter, since they can feed on the hay stored in their burrows. Using the growth-curve criterion overcomes the disadvantages of the static criterion, however, it ignores those animals from the year before that did not reach the estimated maximum juvenile weight. This criterion thus theoretically overestimates the number of juveniles in a given year. But it is also possible that a dynamic criterion underestimates the number of juveniles in a given year. The individuals with the highest growth rates may be missed, so that animals may be considered yearlings (thus coming from the last year), although they were born in the given year. In this section, both criteria are evaluated and compared with respect to their implications on age structure for the observed population, especially the development of yearling numbers. In the following it is expected that all age groups except juveniles should stay constant or decline in numbers. This implies, that mortality is the only important factor influencing the numbers, since immigration, increased mobility, or increased encounter success are neglected. These assumptions seem reasonable, since pikas did not move far during during the study period (Section 2.6, p. 42).

Methods Captured individuals of all 16 capture sessions (Table 2.6, p. 29, from June 2000 to July 2002) were used to study weight structure. Median weight for female, male, and unsexed individuals were calculated and captured individuals were grouped into weight classes diﬀering by 10 g each for plotting. The range for an average adult weight was approximated using individuals recaptured during a time period exceeding 3 days. Growth of these individuals was followed until their last capture. They were grouped

58 4. LIFE HISTORY 4.1. AGE STRUCTURE AND LIFE TABLES

according to the initial weight at their first capture. Groups were selected by trial and error to identify a weight range, where most of the individuals within a group stayed within this range for most of the time. Minimum winter weight was estimated by using all individuals captured first during a capture session in winter. “Winter” was delimited by 1st September and 1st May. Seasons are discussed in a later part of this section, see page 69. The dynamic weight limit for juvenile weight was based on juvenile growth and an approximation of the onset of reproduction for each of the 3 summers. Maximal juvenile growth was estimated using individuals recaptured within more than 3 days and less than 150 days. Animals captured during two additional sample sessions on two burrows between May and June in 2001 were used in addition to the regular 16 capture sessions to gain more information on juvenile weight development. For each individual, the weight differences between two capture events were calculated. The maximum weight difference observed in one indivudual was then used for further analysis of growth rates. From all individuals those showing the highest growth rates were used to construct maximal possible weight gain. Growth rates were not extrapolated for periods of times longer than observed during fieldwork. The weight of the heaviest juvenile captured in May 2001 was used as initial weight for a theoretic individual growing with this maximum growth rate. Growth was assumed to be terminated by the time this theoretic individual reached the lower limit of the adult average weight. To approximate onset of growth, weight structures from the same time periods of the different years were compared. Another information used was an observation during field work: in 2002, all animals captured for the first time showed no signs of having changed their fur from winter to summer. In contrast, individuals having been captured the year before showed irregular patterns of lighter and darker hair on their body. It was thus assumed that all of the new captures in 2002 were young of this year. The curve of maximal growth and therefore the onset of growth was adjusted accordingly. Since most of the data came from 2001, there was no change in the beginning of growth assumed for 2001. The two criteria were evaluated by looking at the development of yearling numbers. Weight at first capture was used to assort captured individuals to a birth year for each of the two criteria. The criteria were only used in the summer months from 1st May to 1st September. Newly captured individuals in the remaining months (i.e. during winter), were always assumed to be born in the summer before. To evaluate the two criteria and to decide upon which to use for further studies, their implication on the age structure was compared using the encounter sessions with parallel capture and observation (November 2000 to July 2002, see Table 2.6, p. 29). All individuals known to be alive were included into the data set, even if they were not observed in a given encounter session. Individuals were known to be alive if they were observed before or after the given capture session.

Results Weight class composition Weight of the captured animals ranged from 27 g to 281 g (Figure 4.1). From 17th July 2000 to April 2001 pika weight ranged between 160 g and 281 g. Medians for this period ranged between 185 g and 213.5 g (Table 4.1). In April 2001 there was one animal considerably heavier than the rest (a female weighing 281 g). Without this animal, weight ranged up to 250 g during this time period.

59 4.1. AGE STRUCTURE AND LIFE TABLES 4. LIFE HISTORY 8 01−23 female 4 male

0 unsexed 8 03−23 4 0 8 04−25 4 0 8 05−22 4 0 8 07−06 06−24 07−01 4 0 8 07−17 07−27 07−07 4 0 8 08−23 08−22 4 0 Number of individuals within weight class 8 09−28 09−23 4 0 8 10−26 4 0 8 11−28 4 0

50 150 250 50 150 250 50 150 250 2000 2001 2002 Weight classes [g]

Figure 4.1: Weight structure of the individuals captured from July 2000 to July 2002. Num- bers in the right corners of each plot are month and day of the capture session. Individuals were assorted to weight classes of 10 g. The years 2000 to 2002 are represented in the columns, while rows represent capture sessions in the same time of the year.

60 4. LIFE HISTORY 4.1. AGE STRUCTURE AND LIFE TABLES

Table 4.1: Median weights for the individuals captured during the 16 capture sessions described in Table 2.6 on page 29. “n” gives the number of individuals captured.

Capture occasion All individuals Female Male Unsexed weight [g] n weight [g] n weight [g] n weight [g] n 1 2000.07.06 170.0 15 177.5 10 210.0 2 155 3 2 2000.07.17 181.0 10 185.0 9 - 0 165 1 3 2000.08.23 213.5 8 208.5 6 223.5 2 - 0 4 2000.09.28 213.5 14 210.0 13 237.0 1 - 0 5 2000.10.26 194.5 16 189.0 12 218.5 4 - 0 6 2000.11.28 205.0 15 197.5 10 225.0 5 - 0 7 2001.01.23 190.0 7 191.0 6 178.0 1 - 0 8 2001.03.23 210.0 9 191.5 6 225.0 3 - 0 9 2001.04.25 211.0 11 211.0 9 205.5 2 - 0 10 2001.05.22 42.0 18 42.0 13 129.0 4 38 1 11 2001.06.24 91.0 54 95.5 34 85.0 13 66 7 12 2001.07.27 180.0 31 174.0 19 211.0 10 159 2 13 2001.08.22 184.5 18 184.5 14 205.0 4 - 0 14 2001.09.23 192.0 14 192.0 11 216.0 3 - 0 15 2002.07.01 142.5 22 139.0 19 181.0 3 - 0 16 2002.07.07 154.5 14 149.5 12 182.0 2 - 0

This “winter” range was observed again in August 2001. Medians for the months from the middle of July 2000 to April 2001 ranged between 184.5 g and 213.5 g. The highest medians were reached in August and September 2000. Median weight in Au- gust and September 2000 was 213.5 g, while it was 184.5 g and 192 g in August and September 2001. The heaviest individual caught (281 g) was a female captured in April 2001. Male median weight was higher than female median weight in 13 capture sessions, but in January, April, and June 2001 female median weight surpassed male median weight. In five sessions there were animals whose sex couldn’t be identified. These unsexed animals always had a median weight smaller than that of male or female individuals (Table 4.1). In May 2001 the first juveniles appeared (Figure 4.1). The individual with the smallest weight was captured in May 2001, weighing 27 g. In this month juveniles could be easily discriminated from adults, since the distribution of weights was bimodal. A difference of 152 g separated the heaviest juvenile from the lightest adult. Juvenile weight ranged between 27 g and 53 g, while adult weight ranged between 205 and 224 g. One month later the gap between juveniles and adults was not evident anymore. Still, weight of all captured individuals included very small animals, ranging from 33 to 229 g. By August 2001 the “winter” weight range was reached again.

Average weight range for adults For 69 individuals (50 females and 19 males) weight development could be followed for more than 3 days. 12 individuals, all of them females, could be observed for more than 300 days. The longest period of time

61 4.1. AGE STRUCTURE AND LIFE TABLES 4. LIFE HISTORY 300 200 100 0 300 200 100 0 300 200

100 female Individual growth from weight at initial capture [g] male 0

0 5 10 15 0 5 10 15 Time passed since initial capture [30d]

Figure 4.2: Weight development of captured individuals. Individuals are separated into 6 groups depending on their weight at initial capture. Weight limits separating groups are indicated as straight lines at 100 g, 160 g, 180 g, 200 g, and 220 g. Animals weighing less than 180 g at ﬁrst capture mostly gained weight and individuals weighing more than 200 g mostly lost weight.

over which weight development could be followed was 444.5 days. (This individual was also the ﬁrst one captured and marked during this study!) Six groups were deﬁned depending on weight at initial capture. Weight limits were 100, 160, 180, 200, and 220 g. Most of the individuals with initial weights between 180 and 200 g showed constant weight (Figure 4.2). This group consisted of 4 females and 2 males. Most of the individuals with initial weights below 180 g gained weight. This group consisted of 36 females and 8 males. Accordingly, most of the individuals with initial weights above 200 g lost weight again. This group consisted of 10 females and 9 males. Adult weight for pikas was therefore supposed to be between 180 g and 200 g. As approximation, 190 g was used for further calculations on mean weight of Ochotona pallasi. Individuals having reached 180 g were considered adult. Although males tended to have higher weights than females, those with high weights lost weight again. Additionally there were only few males and these could not be followed for longer time periods. The average weight range for O. pallasi is thus largely based on female individuals.

62 4. LIFE HISTORY 4.1. AGE STRUCTURE AND LIFE TABLES

Figure 4.3: Maximum female weight diﬀerences observed 150 male in the captured individuals. The line connects the individuals showing the highest

100 growth rates. Mind that this diﬀerence has to be added to the weight of a

50 growing individual. For the Weight Difference [g] heaviest juvenile captured in May 2001 weighing 53 g

0 a diﬀerence of 127 g is necessary to reach the adult 0 2 4 6 8 10 12 14 16 18 weight of 180 g. This would be reached within 6.7 weeks Time Difference [weeks] (47 days).

Table 4.2: The four individuals growing fastest during the given time intervals. These individuals constitute the ascending part of the convex hull describing maximum observed growth (Figure 4.3). ∆ weight gives the diﬀerence in weight and ∆ time the diﬀerence in time between the two capture dates given in date.

Ind. ∆ weight [g] ∆ time [d] Weight [g] Date Sex 056 61.0 14.00 49.0 – 110.0 22.05.01 – 05.06.01 f 055 108.0 33.00 41.0 – 149.0 22.05.01 – 24.06.01 f 061 152.0 92.04 42.0 – 186.0 23.05.01 – 23.08.01 f 057 164.5 123.04 41.5 – 206.0 22.05.01 – 22.09.01 f

Winter-weight criterion From 1st September 2000 to 1st May 2001, there were 72 observations of 36 animals captured between one and 5 times. Weight ranged from 160 g to 281 g, median weight was 205 g. Minimum winter weight was thus 160 g. This is 20 g less than the lower limit for average adult weight of a pika.

Growth-curve criterion For determining the dynamic weight limit, information on juvenile growth rates and dates for the beginning of reproduction were combined. Maximum weight differences for recaptures within 3 and 150 days could be calculated for 37 female and 13 male individuals. The animals with the highest growth grew 61 g in 14 days, 108 g in 33 days, 152 g in 64 days, and 164.5 g in 123.04 days (Figure 4.3, Table 4.2). The heaviest juvenile of the capture session on 23rd May 2001 weighed 53 g. The difference to the approximate adult weight thus amounts to 127 g. This can be reached after 47 days (6.7 weeks) by the hypothetical fast-growing juvenile. Different pieces of information could be used to estimate a date for the onset of growth in the three summers. For 2001, onset of growth was indicated by the heaviest juvenile captured in May, since this individual was used for the construction of the growth curve. Comparing median weights for the three years, median weight at the end of June and the beginning of July was lower in the drought year 2001 (91 g) than in the other years (Table 4.1). In 2002 it was 142.5 g and in 2000 it was 170 g in the beginning of July. In August and September 2001 median weights (184.5, 192.0 g) were

63 4.1. AGE STRUCTURE AND LIFE TABLES 4. LIFE HISTORY

2000 250 y 100 0

2001 250 y 100 0 Figure 4.4: Juveniles were discriminated from adults using weight at ﬁrst capture. 2002

The two criteria used were the winter-weight Weight at first capture [g] criterion using the minimum observed winter 250 weight (dotted line), and the growth-curve y criterion based on growth rates of captured 100 individuals (black curve). The drought sum- 0 mer of 2001 was assumed to have a delayed May Jun Jul Aug Sep onset of reproduction of 14 days as opposed to the other two summers. Date at first capture

lower than those in 2002 (213.5 g). It was thus assumed that reproduction was delayed in the drought summer relative to the other summers. Allowing for an onset of growth 14 days earlier than in 2001 ensured that all of the newly captured individuals in 2002 were considered juvenile (Figure 4.4). This date was used for the years 2000 and 2002. Since in 2001 there was a 53 g juvenile by 22nd May, it was assumed that in 2000 and 2002 there could be a 53 g juvenile by May 8th.

Resulting cohorts For 153 animals the year of birth could be estimated using the two criteria. 129 of them were captured for the ﬁrst time in the summer months between May and September 15th; 24 in the year 2000, 80 in 2001, and 25 in 2002. Figure 4.4 shows the weight of the newly captured individuals together with the threshold lines resulting from the winter-weight criterion or the growth-curve criterion. The threshold line separates “juveniles” from “adults”.

Comparing the winter-weight criterion Table 4.3: Number of individuals as- and the growth-rate criterion In each sorted to the cohorts of 1999 to 2002 year the winter-weight criterion suggested based on the winter-weight (winter) cri- fewer juveniles. According to the winter- terion and the growth-curve (growth) cri- weight criterion there were 28 juveniles and terion to separate juveniles from adults. 18 adults in the year 2000, 69 juveniles 1999 2000 2001 2002 and 13 adults in 2001, and 19 juveniles and 7 adults in 2002. At the same time the winter 18 41 75 19 growth-rate criterion resulted in 37 juveniles growth 9 43 76 25 and 9 adults for the year 2000, 76 juveniles and 6 adults for 2001, and 25 juveniles and no adults for 2002. The resulting cohort strengths are listed in Table 4.3. Since individuals classiﬁed as adult in a given summer

64 4. LIFE HISTORY 4.1. AGE STRUCTURE AND LIFE TABLES

30 01−23 female

15 male unsexed 0 30 15 0

30 04−25 15 0

30 05−26 15 0

50 06−28 07−02 20 0

30 07−30 07−11 15 0

30 08−25 15 0

30 09−25 15 0 Number of animals within cohorts for winter and growth criteria 30 15 0

30 11−28 15 0 9 0 1 2 9 0 1 2 9 0 1 2 9 0 1 2 9 0 1 2 9 0 1 2 winter growth winter growth winter growth 2000 2001 2002

Figure 4.5: Age class composition predicted by the winter-weight criterion (winter) and the growth-curve criterion (growth). Both criteria were applied to the pooled data from the 10 joint observation and capture sessions from November 2000 to July 2002. Note the diﬀerent scales for the months of June 2001 and July 2002. Grey grid lines are always in intervals of 10 animals. must have been born the year before (or earlier), the strength of one year’s cohort is the sum of the juveniles from that year and the newly captured adults from the following year. Thus, the 41 animals belonging to the 2000 cohort according to the winter-weight criterion consist of the 28 juveniles from the year 2000 and the 13 adults of the year 2001. Taking the winter-weight criterion resulted in more older animals for each sampling

65 4.1. AGE STRUCTURE AND LIFE TABLES 4. LIFE HISTORY

session than taking the growth-curve criterion (Figure 4.5, p. 65). Compared to June there was an increase of 8 animals for the yearlings at the end of July. Most of the new-comers were males, reducing the sex ratio from 1.5 females/males to 0.9 in July and 0.7 in August. In August the winter-weight criterion predicted more adult animals than juveniles. Sex ratio of the yearlings was reduced also 2002 by an increase of male individuals. Taking the growth-curve criterion estimated more younger animals. In that case, numbers of yearlings did not increase in Summer 2001, but the sex ratio of yearlings changed from 2.8 to 0.9 females/males, because number of females decreased and the number of males stayed constant. The growth-curve criterion thus performs better in grouping animals so that yearlings did not show increasing numbers. For further analysis this criterion was used to assort individuals to a birth year.

Discussion Using a dynamic weight limit resulted in more consistent estimates for age than using minimum winter weight. Male one-year-old individuals did not increase in numbers in the summer and all the individuals without signs of changing fur could be considered juvenile. But even the dynamic limit seems to be too conservative for males. There are at least three reasons, why male one-year-old individuals show constant numbers in summer, while females do not: (1) higher survival, (2) higher mobility, so that more males are captured, although they had same survival rates, and (3) higher growth rates so that young of the year were confounded with male yearlings. Since male pikas invest less into reproduction than females (Smith and Gao, 1991), male survival may be higher during the reproductive season. In populations of Ochotona daurica observed by Tsendzhav (1977) the ratio of females to males changes in their third year of age. Up to year 2, there are more females, while from year 3 to 4, there are more males. Male mobility in the present study was higher in early summer (March to May, see Section 2.6, p. 42), but not in the last summer months. Higher male mobility can thus be a reason for higher numbers of detected individuals only in early summer, but not in July. The most simple explanation however is that young of the year and yearlings were confounded even when using the dynamic weight limit, since males showed, at least for some time, higher body weight than females (Figure 4.2, p. 62). Reviews on pikas state that there is no sexual dimorphism in pika weight (Smith, 1988; Smith et al., 1990). Puget and Gouarderes (1974) even show that weight development of juveniles of both sexes was similar from birth to 19 weeks of age for individuals of Ochotona rufescens held in captivity. The same is true of O. curzoniae held in captivity (Ye and Liang, 1989, and Figure 4.6, p. 67). In contrast to this, this study indicates that male, free living Mongolian pikas might have a diﬀerent growth pattern than females. But although male pikas showed higher body weight than females at ﬁrst capture, most of these individuals lost weight again (Figure 4.2, p. 62). There may be a similar body weight of adult female and male pikas, but young male pikas seem to gain weight faster and reach a higher body weight for a certain time. In the observed population, juveniles entered traps in May, weighing between 27 and 53 g (Figure 4.1, p. 60). Relying on the growth rates given for captive Ochotona rufescens by Puget and Gouarderes (1974) and Ye and Liang (1989) for O. curzoniae, these animals could have been between 8 and 17 days old (Figure 4.6, p. 67). O. rufescens is somewhat heavier than Ochotona pallasi pricei, Puget states a mean weight of 225 g for 9 females and a maximum weight of 250 g, while the average weight of Ochotona

66 4. LIFE HISTORY 4.1. AGE STRUCTURE AND LIFE TABLES

Figure 4.6: Growth curves of

200 Ochotona pallasi pricei as estimated in this study, for O. rufescens

150 (Puget and Gouarderes, 1974) and O. curzoniae (Ye and Liang, 1989). Both were laboratory studies and 100

Weight [g] the did not detect a diﬀerence in O. p. pricei male and female growth. Early ju- 50 O. rufescens venile growth was assumed to be O. curzoniae similar for the present study. For

0 Ochotona p. pricei the heaviest juvenile captured in May 2001 weighed 0 20 40 60 80 100 120 140 53 g. This weight was reached between 14 and 16 days of age in the Days from birth other species. p. pricei was supposed to range between 180 to 200 g in the present study, at least in a drought year (Figure 4.2, p. 62). The gestation period in Ochotona rufescens is 24 days, being in accordance with a probable gravid female captured in April in the observed population. Already by August Ochotona p. pricei reached the same population weight range as in the previous winter, while the captive animals of O. rufescens do not reach adult weight within 19 weeks, which would be about August beginning in May. In the optimum scenario during the drought summer, an individual weighing already 53 g in May could reach the adult weight range of 180 to 200 g within 47 days, less than 2 months (Figure 4.3, p. 63). Ochotona pallasi pricei thus grows faster than O. rufescens or O. curzoniae (Figure 4.6). In contrast to these findings, Krylova (1973) reports that young of the year and yearlings of Ochotona pallasi pricei have lower weights in summer than two to four year old animals. Age was determined by counting adhesion lines in periosteal bone of the lower jaw (see also Millar and Zwickel, 1972). At least some of these yearlings would have been considered young of the year taking the weight criteria used here. Minimum weight measured by Krylova (1973) in August (113 g) is lower than the minimum weights measured in this study (150 g), where the weight range in August 2001 resembled that of the previous winter. Judging from this study, it is not plausible that Ochotona pallasi pricei needs more than 3 months to reach adult weight, while the data from Krylova (1973) imply growth continuing in the second summer. Weight of two to four year old individuals ranges from 202 to 277 g in her study, which is in accordance with the present study (Figure 4.1, p. 60). The dramatic seasonal differences in a continental climate imply a regular pattern for the onset of reproduction. Since winters are very cold and summers are short, pika reproduction has to be synchronised. However, in the present study there are indica- tions of a variable onset of reproduction. Median weight was lower at the end of June and beginning of July in the drought year, the same was true of August. In a four year study Wang (1990) shows that mating seasons begin between the end of March to the second 10-day period of April for Ochotona curzoniae. Lagomorphs are known for their ability to adapt fecundity to climatic conditions (Chapman and Flux, 1990a). Varying intraspecific litter sizes result in almost constant annual production of young per female across latitudinal gradients in New World rabbits (Sylvilagus and Romero-

67 4.1. AGE STRUCTURE AND LIFE TABLES 4. LIFE HISTORY

lagus) and hares (Lepus). Varying interspecific gestation periods across latitudes lead to optimal conditions to produce a maximum number of young in rabbits. Although lagomorphs are not as diverse as rodents, they make up the base of many predator-prey systems because of their high abundance in ecosystems. Thus, even though reproductive activity can rely on very regular changes in temperature from winter to summer, reproduction of the Mongolian pika does not seem to be triggered simply by temperature, but also shows variability related to food availability. Nevertheless, reproduction takes place during the short summer months, while no reproduction was observed during the winter. This makes it possible to model pika population dynamics in discrete time steps, using one year or half a year, depending on the need to account for the different number of individuals in winter and summer. Highest median weights were reached in August and September, both in the drought year and in the year before (Table 4.1, p. 61). This is in accordance with the end of the vegetation period (Retzer, 2004). After this, pikas have to rely on the plants they have collected and stored and on the remaining standing crop. Pikas reduce standing crop even in winter (Retzer, 2004), so they do not exclusively feed on the plants they have collected during summer. During winter, there was no large reduction in weight. In contrast to livestock or other non-caching organisms, pika do not have to rely on fat reserves to survive the winter. The relative constancy in body weight during the winter nicely illustrates this. Although Stubbe (1971) stated that it is impossible to distinguish juveniles from adult pikas by inspecting their fur, this study indicated that it may be reasonable to rely on fur characteristics. The study cited above refers to data collected in August, when all animals, including the young of the year, were already changing to their winter pelage. In July however, the molt might not have begun yet. The only animals with signs of fur change in the present study were those that still lost the hair from the preceding winter, thus, adult individuals. But even later in the year, the change from summer to winter fur was not as evident as the change from winter to summer fur (observations from the year 2001). At least until August it might be a good first estimator for age to look at signs of fur change.

68 4. LIFE HISTORY 4.1. AGE STRUCTURE AND LIFE TABLES

4.1.2 Seasonal development of age structure When modelling population dynamics one has to decide at which time scale the system should be observed. Although trapping took place in approximately monthly intervals, this time scale may be unnecessarily ﬁne for modelling. Therefore, this section looks at patterns in survival of the observed population. It asks, if certain months in the year provide equivalent information and thus data can be summarised in seasons.

Methods Age was determined based on the growth-curve criterion described above. Three data sets from the observed population were compared: (1) data on the individuals captured only on the smaller part of the study site, where capture began already in July 2002, (2) data from the enlarged study site beginning in September 2002, and (3) pooled data from combining capture and observation (Table 2.6, p. 29). The ﬁrst data set has the lowest number of individuals, while the third data set has the highest number of individuals. On the other hand the ﬁrst data set spans the longest period of time including three summers, while the third data set spans the shortest time. Individuals known to be alive, i.e. those encountered before, after, or during an encounter session were included in a given session. Individuals were grouped into cohorts and sexes. The number of individuals at each encounter session was calculated. To assess an age structure of a typical population of Ochotona pallasi pricei using data from the reproductive season might be misleading, since many juveniles may die until winter. Winter data were available only for 2000/2001 though, so that data for the years 2001 and 2002 had to be taken from September and July respectively.

Results Pooled data from capture and observation showed little reduction in numbers from November 2000 to April 2001 for both the 1999 and the 2000 cohorts (Figure 4.7). Compared to the capture data from the enlarged study site (Figure 4.8), pooled data up to April 2001 still outnumbered the individuals captured in September and Octo- ber 2000 for the cohort of 2000. The numbers were reduced by one individual for the cohort of 1999. The pattern was the same for the cohort of 1999 on the small part of the study site (Figure 4.9), where numbers were reduced by one animal from September to April. Both data sets on captured individuals showed a decline in numbers from September to April for the cohort of 2000. Pooled data from capture and observation showed a decline in numbers from May 2001 to August 2001 for the cohort of 2000. The older individuals (cohort of 1999) were again reduced by one individual. Males of the 2000 cohort were not reduced, but this may be an artefact of separating juveniles from adults based on weight (see Section 4.1.1, p. 58). From August to September numbers stayed constant for pooled data for both cohorts (1999 and 2000). For capture data alone the decline in numbers did not stop in September 2001 for the 1999 cohort and the 2000 one. From the 2000 cohort, there were always fewer animals detected by capture alone than by pooling capture and observation. All three data sets showed the emergence of juveniles in May 2001, an increase up to June and a decline in numbers up to September. The rate of decline was smallest for the period between August and September 2001. There was a decline also of juveniles

69 4.1. AGE STRUCTURE AND LIFE TABLES 4. LIFE HISTORY 6

4 Cohort 1999 2 0

Cohort 2000 20 10 0

Cohort 2001 40 20 0

female Cohort 2002 20

Number of individuals encountered male

10 unsexed 0

May Sep Jan May Sep Jan May 2001 2002

Figure 4.7: Individuals encountered by capture and observation, separated for the cohorts. Note that the y-axes use diﬀerent scales. Survival of adults was relatively high compared to juvenile survival. A “+” indicates a sample session. 6

4 Cohort 1999 2 0

Cohort 2000 20 10 0

Cohort 2001 40 20 0

female Cohort 2002 20

Number of individuals encountered male

10 unsexed 0

May Sep Jan May Sep Jan May 2001 2002

Figure 4.8: Individuals encountered by capture only, separated for the cohorts. Note that the y-axes use diﬀerent scales. A “+” indicates a sample session.

70 4. LIFE HISTORY 4.1. AGE STRUCTURE AND LIFE TABLES 6

4 Cohort 1999 2 0

Cohort 2000 20 10 0

Cohort 2001 40 20 0

female Cohort 2002 20

Number of individuals encountered male

10 unsexed 0

May Sep Jan May Sep Jan May 2001 2002

Figure 4.9: Individuals encountered by capture on the small part of the study site, separated for the cohorts. Note that the y-axes use diﬀerent scales. A “+” indicates a sample session.

Table 4.4: Numbers and per- Numbers Percentage [%] centages of female individuals in with age 0–3 [a] with age 0–3 [a] each age group at the end of sum- Date MNA 0 1 2 3 sum 0 1 2 3 mer in each of the years 2000– 2000 31 16 3 – – 19 84 16 – – 2002. Data are based on pool- 2001 35 15 8 2 – 25 60 32 8 – ing information from capture and 2002 29 15 6 2 0 23 65 26 9 0 observation from January 2001, September 2001, and July 2002.

of 2002, while the numbers of adults stayed constant in all data sets. Pooled data detected two-year-old individuals in 2002, while they were absent in capture data. The age structure summarised for the three years showed that the percentage of juveniles was highest in the winter of 2000/2001, when 84 % of the marked females were considered young of the year (Data from January 2001). After summer 2001 the percentage of juveniles was 60 % and in summer 2002 it was 65 % (Table 4.4). Contrasting young of the year and older individuals for the three summers did not show significant differences, though: a χ2 based comparison of deviances using a generalised linear model with Poisson errors did not find significant differences between the years (Difference in deviance = 0.4, df = 2, p = 0.83), while differences based on age were highly significant (Difference in deviance = 30.9, df = 2, p < 0.001). Based on the median the juveniles accounted for 67 % in the average age structure of the whole study period. Two-year-old individuals made up 8 and 9 % of the population relying on the numbers of the summers of 2001 and 2002. This results in 67 % juveniles, 24 % one-year-old individuals, and 9 % two-year-old or older individuals.

71 4.1. AGE STRUCTURE AND LIFE TABLES 4. LIFE HISTORY

Discussion

Survival and reproduction patterns allow for the distinction of two seasons in the Mongolian pika: the “summer” season from May to August and the “winter” season from September to April. The summer-season is characterised by reproduction and low survival of juveniles. Adult survival also is lower in this part of the year. In the “winter” season, survival is high for all age groups and there is no reproduction. Although capture data showed a decline for the 2000 cohort in this period, pooled data always detected more individuals. Numbers of males from the cohort 2000 did not decline in the summer of 2001. This probably was an artefact, since the weight limit separating juveniles from adults was derived by using mainly female individuals (Section 4.1.1, p. 58). Using this pattern, the year can be divided into three parts of equal length, namely 4 months, starting on 1st May, 1st September, and the 1st January. Two of these parts belong to the winter season, one to the summer season. Since reproduction only takes place in the summer, pikas can be viewed as pulse breeders and dynamics can be described by discrete steps of 4 months each. The growth season of vegetation corresponds with this division: growth starts during April and highest standing crop is reached in August (Lavrenko et al., 1993; Retzer, 2004). In the observed population, younger individuals were always more numerous than older individuals (Table 4.4, p. 71). Young of the year were especially numerous after summer 2000. There are several drawbacks when comparing the numbers presented in Table 4.4. Since age was assorted to animals only during the summer months (May to September), newly captured animals in October 2000 to January 2001 were considered to be born in 2000. The number of marked individuals increased with becoming familiar to the study site and the observed animals. It is thus possible that some of the animals captured after summer 2000 were indeed not from 2000 but older. Age structure will be further discussed in the concluding Section 4.5, p. 101. It is possible that juvenile survival was lower in the year of the drought and in the following year, than in the more favourable year 2000. The numbers in the year 2002 are similar to those of 2001, but they were sampled already two months earlier. Although juvenile ratio was higher in July 2002 than in September 2001, it could have decreased again if data were also taken in September 2002. On the other hand, survival may also have been high from July 2002 to the next summer, because more food was available and fewer individuals were present in 2002. In July of that year already all the individuals had a burrow to live in, which could have increased their probability to survive until the next summer. This is only one of many possible scenarios. Numbers were definitely lower in summer 2002 than in the drought summer of 2001, indicating a response to the drought with the delay of one year. The animals born in summer 2000 always accounted for the highest percentages in their age groups, although this pattern could have been produced by chance alone. They made up 84 % in 2000, 32 % in 2001, and 9 % in 2002. This may be caused by the favourable conditions they experienced in 2000, indicating that survival might not only be effected by the present, but also by past weather conditions. Physiological condition, possession of a territory, or both may have been influenced by past conditions. Juveniles developing in a favourable year may have the chance to grow stronger and healthier than juveniles developing under drought conditions. Median weights in August and September were lower in the drought year than in the preceding year, indicating that

72 4. LIFE HISTORY 4.1. AGE STRUCTURE AND LIFE TABLES

growth was influenced by the drought (Table 4.1, p. 61). The individuals surviving the winter into the drought year may have selected the best territories before the onset of reproduction and have achieved high social ranks (Holst et al., 1999), while the juveniles may have not been able to establish themselves in good territories at the end of the summer. Such cohort effects are documented for other species (Bjørnstad et al., 2004; Stenseth et al., 2002) and will be one of the factors influencing survival rates tested for the present study in Section 4.2.2 on page 81.

73 4.1. AGE STRUCTURE AND LIFE TABLES 4. LIFE HISTORY

4.1.3 Cohort life tables Life tables (Krebs, 1985) were calculated based on cohort and season as delimited above. For the life tables, only the number of individuals at a given age and the survival rate from a given age group to the following were given. The year was divided according to the seasons delimited in the preceding section: it consisted of three parts of 4 months each, starting on 1st January. The number of individuals in a given age group was derived from all individuals encountered by capture or observation during that period of time. Data from all capture and observation sessions, including additional data from capturing on two burrows in the beginning of June 2001 and from extra observation around the study site in September 2001 and July 2002 (Table 2.6, p. 29) were used. Individuals encountered before and after the given time period were included. Survival rates from one age group to the next were given in units of 30 days. To make survival of juveniles comparable, a total of 180 juveniles (60 female, 60 male, 60 unsexed) weaned per summer were assumed. Survival of juveniles in the summer of 2002 was calculated by assuming that three months of the summer had passed, since capture and observation did not take place within the following part of the year, but only within the same summer (in July 2002). Survival rates (fa) in units of 30 days for the given age a were calculated according to the following equation, where na is the number of individuals at a given age, while na+1 is the number of individuals present in the next following time step for which information was available. The number of days separating these time steps is given by t. 30 30 t na+1 na+1 f = t = (4.1) a n n r a a Separate life tables are given for the cohorts 1999 to 2002, for females, males, and unsexed individuals, and for individuals from early and late litter in 2001. Early and late litter were delimited by weight at the time of ﬁrst capture. All juveniles captured in May 2001 and those captured in the beginning of June weighing more than 65 g were considered early born, while those captured in the end of June weighing less than 65 g and those captured in July weighing less or equal than 85 g were considered late born. For comparison with weight structure see Figure 4.1 on page 60. Seventeen individuals were assorted to the group of early litter, while 11 individuals were assorted to late litter. To compare numbers of individuals in the resulting groups, χ2-tests were used with two-way tables, whereas generalised linear models with Poisson errors were used with more than two explaining factors.

Results For each of the cohorts a diﬀerent time frame could be observed. For the 1999 cohort the age from one to three years, for the 2000 cohort the time interval from birth to two years of age, for the 2001 cohort from birth to one year of age, and for the 2002 cohort only the summer of birth. 30 day survival rates from birth to the next observed age group declined with the years. Time intervals were one year for the 1999 cohort, 4 months for the cohorts of 2000 and 2001, and 3 months for the cohort of 2002 (Table 4.5, p. 75). The lowest observed survival rate was zero for the 2.3 year old animals (ﬁrst part of their third winter). These three animals did not survive their third winter. Thus, 2.3 years was the longest life span observed in the population of the Mongolian pika.

74 4. LIFE HISTORY 4.1. AGE STRUCTURE AND LIFE TABLES

Table 4.5: Life table for 1999 2000 2001 2002 the cohorts from 1999 to a n f n f n f n f a a a a 2002. 30 day survival (180) 0.78 (180) 0.68 (180) 0.58 (180) 0.52 rates (fa) were calculated 0 43 76 25 assuming a total of 180 0.3 – 38 0.96 20 0.89 – juveniles being born in 0.7 – 32 0.91 – – each year. Age (a) is 1 9 0.90 22 0.91 8 – given in years, 0 for the 1.3 6 1.00 15 0.85 – – ﬁrst summer (from May 1.7 6 0.96 – – – to September), 0.3 and 0.7 2 5 0.88 4 – – for the two winter parts 2.3 3 0.00 – – – from September to Jan- 2.7 – – – – uary and from January to 3 0 – – – May respectively. Dashes (–) indicate that no data are available.

Highest survival rate was 1.0 for the 1.7 year old animals (second part of their second winter). None of the remaining survival rates was below 0.85. Over winter survival of the yearlings was higher for the 1999 cohort (from 2000 to 2001), than for the 2000 cohort (from 2001 to 2002). This difference was not significant using a χ2-test (χ2 = 1.1, df=1, p=0.284). Additionally, sampling intensity was much higher for the 1999 cohort, since exhaustive data exist for winter 2000/2001 and only one sample session in September 2001 for the winter 2001/2002. Comparing the 1999 cohort with that of 2000 including the 180 assumed new born individuals, there were more yearlings in the 2000 than in the 1999 cohort. This difference nearly reached significance (χ2 = 5.2, df=2, p=0.074). Comparing the 2000 cohort with that of 2001, over winter survival of the new born animals from the first part of the first winter (age 0.3) to the following summer (age 1) was 0.93 for the 2000 cohort and 0.89 for the 2001 cohort (calculated according to equation 4.1, p. 74). But the differences between the cohorts may have arisen by chance alone (χ2 = 0.3, df=1, p=0.614). Only when including the 180 new born, the differences became significant (χ2 = 9.9, df=2, p=0.007). Thus, assuming an equal number of new born individuals, the cohort from the year 2000 produced more yearlings than the one from the year 2001. Survival of female juveniles was higher than male survival in all cohorts. Survival of unsexed animals was lower than those of the other animals for the first time intervals (Table 4.6, p. 76). Over winter survival for the first winter was higher for females from the 2001 cohort, where no male animal survived the winter. For both females and males from the 2000 cohort survival for the first winter was greater than 0.9. Over winter survival for the second winter was higher for females from the 1999 cohort, while no male survived the second winter. Survival was slightly higher in males than in females for the 2000 cohort. Overall survival up to age two was equal in males and females of the 2000 cohort, 2 females and 2 males survived the second winter. Age was the only factor with a significant effect on survival when analysing the data with a generalised linear model with Poisson errors, holding the other factors (sex and cohort) constant. For this test, only data of male and female individuals for 0,1, and 2 years of age were compared. Two separate tests were performed, thus avoiding missing values: first comparing all cohorts of age 0 to 2, and then comparing all ages without the cohort of 2001. Both models could be reduced by sex and cohort without

75 4.1. AGE STRUCTURE AND LIFE TABLES 4. LIFE HISTORY

Table 4.6: Life tables for the co- Female Male Unsexed horts (Coh.) from 1999 to 2002 sep- Coh. a n f n f n f arating sexes. 30 day survival rates a a a (fa) were calculated assuming a to- 1999 (60) 0.83 (60) 0.76 (60) 0.71 tal of 60 juveniles beeing born in each 0 – – – year in female, male, and unsexed in- 0.3 – – – dividuals. Age (a) is given in years, 0.7 – – – 0 for the first summer (from May to 1 6 0.96 2 0.00 1 1.00 September), 0.3 and 0.7 for the two 1.3 5 0.95 0 1 1.00 winter parts from September to Jan- 1.7 4 1.00 0 1 1.00 uary and from January to May re- 2 4 0.93 0 1 0.00 spectively. Dashes (–) indicate that 2.3 3 0.00 0 0 no data are available. 2.7 – – – 3 0 0 0 2000 (60) 0.77 (60) 0.72 (60) 0.36 0 22 16 5 0.3 21 0.98 16 0.93 1 1.00 0.7 19 0.91 12 0.93 1 0.00 1 13 0.89 9 0.94 0 1.3 8 0.84 7 0.86 0 1.7 – – – 2 2 2 0 2001 (60) 0.73 (60) 0.48 (60) 0.00 0 47 18 11 0.3 17 0.91 3 0.00 0 0.7 – – – 1 8 0 0 2002 (60) 0.71 (60) 0.41 (60) 0.00 0 21 4 0 changing the fit, using a χ2 test for the differences in deviance (sex: difference=0.490, df=1, p=0.484, cohort: difference=0.015, df=1, p=0.903, age: difference=294.451, df=1, p<0.001 for the first model and sex: difference=0.52, df=1, p=0.47, cohort: difference=0.71, df=1, p=0.40, age: difference=349.92, df=1, p<0.001 for the second model). All factors will be tested later in this study in a more powerful analysis including recapture probabilities (Section 4.2.2 on page 81). Concerning the survival of early or late born juveniles, more individuals of the early born juveniles survived than did those of the late born ones. None of each group survived up to an age of 2 years (Table 4.7, p. 77). However, this pattern may have arisen by chance alone (χ2=2.3, df=2, p=0.3)

Discussion

The shorter the time period between encounter sessions was, the lower were the survival rates of juveniles. This time period was one year for the cohort of 1999, 4 months in 2000 and 2001, and 3 months in 2002. Mortality thus is high when pikas are very young, possibly even before they can be captured. Delayed eﬀects of climate are important in both marine and terrestrial ecosystems

76 4. LIFE HISTORY 4.1. AGE STRUCTURE AND LIFE TABLES

Table 4.7: Life table for those juveniles born at the be- ﬁrst last ginning (early) and at the end (late) of the reproductive age n f n f period in 2001. 30 day survival rates (f) were calculated 0 17 0.70 11 0.55 for each age class. Age is given in years, 0 for the ﬁrst 0.3 4 0.97 1 0.00 summer (from May to September), 0.3 and 0.7 for the 0.7 – – two winter parts from September to January and from 1 3 0 January to May respectively. Dashes (–) indicate that no data are available.

(Stenseth et al., 2002). Individuals born in a specific year may be larger or smaller than the average, depending on the climatic conditions in the year of birth. Such cohort effects have been reported in both ungulate and fish populations as the result of climatic variability related to the North Atlantic Oscillation (Stenseth et al., 2002). Whenever younger age groups are more affected, the ecological effects of climate are more difficult to detect because of cohort effects (Kaitala and Ranta, 2001). Assuming the same number of new born individuals, the present data supported a cohort effect, with the cohort 2000 having more older animals than the cohort 2001. This may indicate an interaction between cohort and drought year when regarding survival rates. Although the pattern may have arisen by chance alone, juveniles born early in the year showed higher survival than those born later in the year, which would indicate that females were energy-limited in summer 2000. This is plausible, since it was a drought summer. It may also show higher survival for early born juveniles in general, indicating a life history trait of a K-type life history pattern (see Section 2.2 on page 23 and Smith, 1988). Difference in seasonal survival rates are inconsistent. Although the cohort of 1999 showed higher survival in winter, the cohort of 2000 did not show such a pattern. There seems to be an interaction between age and sex for survival. Juvenile male survival was always lower than juvenile female survival, while unsexed individuals showed even lower survival. Adult male survival was also lower than female survival for the 1999 and 2001 cohort, since none of the males survived until the next time step. But in 2000 adult male survival was similar or even higher than female survival. This however may be an artefact, since the weight limit used to assort juveniles and adults did not seem to work as well for males as it did for females (Section 4.1.1, p. 58). On the other hand, higher male survival has been reported for Ochotona daurica by Tsendzhav (1977). Female survival may be lower due to a higher reproductive effort when they reach adult age. The next section uses a more powerful tool to analyse the influence of seasons, sex, cohort, date of birth, drought, and density on survival rates than can be done with data from life tables. For this, recapture probabilities are included in the analysis.

77 4.1. AGE STRUCTURE AND LIFE TABLES 4. LIFE HISTORY

4.1.4 Summary In the year 2001 the first juveniles appeared in May, but the beginning of the reproductive season is variable. The reproductive season in the drought year of 2001 probably began about two weeks later than did those in the other two years. Juveniles appear until the end of July. Ochotona pallasi pricei grows faster than the two other species whose growth was studied, Ochotona curzoniae and O. rufescens. The average adult weight is reached between 180 and 200 g. This can be reached by mid July for an individual weighing 53 g in the end of May, thus within the reproductive season. To separate juveniles from adults the growth-curve criterion accounting for growth and a variable beginning of the reproductive period is better suited than the winter- weight criterion, which uses a static limit. Preliminary analysis of the age structure suggests that the year can be divided into two parts concerning survival and reproduction. While in the reproductive season lasting four months from May to September there are great changes, the structure is static during the months from September to May. These seasons roughly correspond to winter and summer season. The average age structure at the end of the summers 2000 to 2002 was 67 % juveniles, 24 % one-year-old individuals, and 9 % two-year-old or older individuals. Summarising information on individuals in life tables shows that the longest observed lifespan is 2.3 years. Juvenile survival tends to be low especially in the first weeks. Survival rates of adult male and females show an inconsistent pattern, differences are not significant. Although late born juveniles show lower survival than early born juveniles, these differences are not significant based on the life tables.

78 4. LIFE HISTORY 4.2. MODELLING SURVIVAL RATES

ρ ρ 2 11 φ1ρ2 3 101 φ1(1 ρ2)φ2ρ3 φ1 φ2 −

1 1 ρ2 10 φ1(1 ρ2) 1 ρ3 100 φ1(1 ρ2)φ2(1 ρ3) − − − − − 1 φ 1 φ − 1 10 1 φ − 2 100 φ (1 ρ )(1 φ ) − 1 1 − 2 − 2 100 1 φ − 1

Figure 4.10: Example showing the derivation of probabilities for the encounter histories “101” and “100”. The lines show the steps leading to the formulae. The ﬁrst question is, if the animal survives (φ) or not (1 φ), the second whether it is recaptured (ρ) or not (1 ρ). − − These are summed up to result in the probability given in Table 4.8 on the following page.

4.2 Modelling survival rates

Factors controlling survival are crucial when trying to understand the population dynamics of a species (Begon et al., 1996). Of the life history traits introduced in Ta- ble 2.4 on page 24, this section analysis the inﬂuence of age, cohort, sex, season, and density on survival rates using capture-recapture analysis (Section 4.2.2, p. 81). Since individuals cannot be followed closely through time, the exact time of death most often remains unknown. In the previous section there was no distinction between animals not encountered but living and animals not encountered because they had died. Analysis techniques that explicitly account for reencounter rates additional to survival rates were developed to overcome this drawback (Lebreton et al., 1992). Cormack-Jolly-Seber models (White and Burnham, 1999) implemented in modelling environments that use maximum likelihood theory to estimate survival and reencounter parameters have a long history in ecology and are brieﬂy introduced in Section 4.2.1. The concept of two opposing suites of life history traits in the pikas includes the prediction, that individuals closer to the K-type life history strategy invest more energy in present survival than in reproduction. This implies, that less energy is invested in a second litter, leading to lower survival rates for juveniles born later in the summer season (Section 2.2, p. 23). Survival rates of early and late litter are analysed in Section 4.2.3 on page 92.

4.2.1 Modelling background The quantative analysis of survival must be based on capture-recapture (or reencounter) models, which consider reencounter rates as nuisance parameters in addition to the parameter of primary interest (Lebreton et al., 1992). Cormack-Jolly-Seber models (Cormack, 1964; Jolly, 1965; Seber, 1965) as implemented for example in the software MARK consider reencounter rates (White and Burnham, 1999). These models, based on reencountering marked individuals, make the assumption that (1) each individual from a selected group of individuals has the same probability of survival (φt) from time t to t + 1. Furthermore, these models assume (2) that the individuals are identical in their probability of being reencountered at time t + 1 (ρt+1) (Lebreton et al., 1992). The procedure can be illustrated taking four possible encounter histories for a

79 4.2. MODELLING SURVIVAL RATES 4. LIFE HISTORY capture-recapture study with 3 sample sessions (Table 4.8, only those marked at the first session). Encounter histories are sequences of the numbers “0” and “1”. At each sample session, a “0” indicates that an individual was not encountered, while “1” indicates the encounter. Reencounter probabilities can then be estimated with the help of animals, which have not been encountered, although they have been alive. Fig- ure 4.10 on the page before shows the probabilities for the encounter histories “100” and “101”, assuming equal survival and equal reencounter rates for each session. There is only one possibility of how the encounter history “101” can come about: the animal survives from the first to the second sampling session (φ1), it is not captured on the second sampling session (1 ρ ), survives from the second to the third sampling ses- − 2 sion (φ2), and is captured at the third sampling session (ρ3). For the encounter history “100” there are three different possibilities. The sum of these three possibility is the probability of the encounter history “100”, as noted in Table 4.8. The model likelihood is calculated as the product of all the probabilities of all the observed encounter histories. In the case of the present example with one initial capture and marking session and two following capture sessions, the four encounter histories given in Table 4.8 present all possible encounter histories. The likelihood of a given data set based on the assumptions stated above is the product of all realised encounter histories. Since animals with the same encounter histories have the same probability expression, the number of individuals observed with the same encounter history can be used as exponent of the corresponding probability in the likelihood.

N111 L = (φ1ρ2φ2ρ3) N φ (1 ρ )φ ρ 101 1 − 2 2 3 N110 φ1ρ2(1 φ2ρ3) − N100 1 φ1ρ2 φ1(1 ρ2)φ2ρ3 (4.2) − − − Taking the natural logarithm then gives the log likelihood of the model. Arbitrarily assuming that 15 animals have had an encounter history of “111”, 11 an encounter history of “101”, 7 “110”, and 20 “100” (Table 4.8, p. 80), illustrates how log likelihood is calculated for a given data set.

ln (L) = N ln (φ ρ φ ρ ) + N ln φ (1 ρ )φ ρ + 111 · 1 2 2 3 101 · 1 − 2 2 3 N ln φ ρ (1 φ ρ ) + N ln 1 φ ρ φ (1 ρ )φ ρ 110 · 1 2 − 2 3 100 · − 1 2− 1 − 2 2 3 (4.3) = 15 ln (φ ρ φ ρ ) + 10 ln φ (1 ρ )φ ρ + · 1 2 2 3 · 1 − 2 2 3 7 ln φ ρ (1 φ ρ ) + 20 ln 1 φ ρ φ (1 ρ )φ ρ · 1 2 − 2 3 · − 1 2 − 1 − 2 2 3 Table 4.8: Four possible encounter histories His. Probabilities n (His.) for three sample sessions with their respective probabilities based on survival (φ) and 111 φ1ρ2φ2ρ3 15 recapture (ρ) rates, assuming capture and mark- 101 φ (1 ρ )φ ρ 10 1 − 2 2 3 ing on the ﬁrst session and only capture in the 110 φ ρ (1 φ ρ ) 7 1 2 − 2 3 following sessions. Numbers of individuals with 100 1 φ1ρ2 φ1(1 ρ2)φ2ρ3 20 the respective encounter histories (n) are arbi- − − − trarily chosen to illustrate how model likelihood is calculated (Equation 4.3).

80 4. LIFE HISTORY 4.2. MODELLING SURVIVAL RATES

Using a logit link for binary data, equation 4.3 on the facing page is then maximised to have the highest likelihood with the given data. In this case, models are calculated to maximise the likelihood for different values of φ1 2 and ρ2 3. Estimates for − − these parameters are the result of the modelling procedure. Variance is calculated based on the curvature of the likelihood function. It is defined as the negative inverse of the 2nd derivative of the likelihood function at the maximum likelihood estimate (White and Burnham, 1999). Hypotheses can be tested by comparing models with different sets of parameters. The fully parameterised model includes main factors and interactions. This model can then be reduced by omitting interaction terms or reducing factor levels. Models can be compared using analysis of variance only if the models are nested. Non-nested models are compared using information theory: the Akaike Information Criterion corrected for sample size (AICc) as provided by MARK weighs model likelihood and the number of parameters for a given model (Lebreton et al., 1992; White and Burnham, 1999). The currently accepted convention is that models with AICc that differ by <2 are indistinguishable in terms of their fit to the data (Burnham and White, 2002). Before models are compared, the fully parameterised model is subjected to a goodness of fit test by using χ2-tests or bootstrapping techniques. The χ2-tests compare reencounters of groups of individuals with expected reencounters based on the model assumptions of equal survival and reencounter rates. Bootstrapping is done by using the parameter estimates from the maximum likelihood calculation. New data sets are simulated with the estimates, and maximum likelihoods are calculated for these data sets. The distribution of bootstrapped model deviances is then used to evaluate the deviance of the original data set. Deviances measure the difference in fit for a given model and a fully saturated model, with one parameter for every unique encounter history. Both goodness of fit tests can be used to calculate a measure of overdispersion, the variance inflation factor (cˆ). This factor can be used to adjust the AICc through quasi-likelihood, resulting in a QAICc whenever cˆ departs from 1. For bootstrapped data sets cˆ is calculated as the devision of observed deviance with mean simulated deviance (White and Burnham, 1999). Depending on the hypotheses to be tested, an a priori set of models can be selected to compete with each other for the highest QAICc value. For example, effects of the factor age can be tested by first running the fully parameterised model and then running a model with one parameter for all age groups. Normalised QAICc-weights are given as a measure of the of a model’s relative probability of being the best model for the data as compared with alternative models (White and Burnham, 1999).

4.2.2 Survival rates Relying on the distinction between two opposing life history traits for the pikas, an r-type of life history strategy as found in burrowing species like the Mongolian pika is expected to show high juvenile survival rates, low overwinter survival, low adult survival, and no density dependence in survival rates (Smith, 1988; Smith et al., 1990; Table 2.4, p. 24). The previous sections (especially Section 4.1.3, p. 74) already indicated a diﬀerent pattern for the observed population of Ochotona p. pricei. Juvenile survival rate seemed to be low, while adult survival was high. Females and males showed diﬀerent patterns of survival, while there were inconsistent patterns concerning overwinter survival. There

81 4.2. MODELLING SURVIVAL RATES 4. LIFE HISTORY

seemed to be interactions between several factors. For example, the drought year affected juveniles more than adults. There may also be an influence dependent on the time of birth, both in terms of years and within the summer season. The cohort from 2000 seemed to have higher survival rates through the years, as have animals born early in the year. In the following analysis, the effects of sex, cohort, time of sample session, and density will be tested using capture-recapture analyses as described above (Section 4.2.1, p. 79). The effect of time of birth within a season will be tested in Section 4.2.3 on page 92. The effect of sex, age and density are tested for recapture rates. Male movement patterns are different from female movement patterns, especially in the summer months (Sections 2.6, p. 42 and Chapter 3, p. 47). Age may effect recapture rates, because juveniles may be more explorative than older animals, especially when they are searching for own territories. Under high density conditions there may be more need for hiding places and thus traps may be better accepted during this time. More factors were included into the survival models. Here, the effects of sex, cohort, season, drought year, age, and density were tested.

Methods Two data sets were used to estimate recapture and survival rates: the first using information from cap- Table 4.9: Number of animals ture and observation, starting in November 2000 for each cohort and each sex used and ending in July 2002. This data set has the for modelling recapture and survival rates. Numbers are given for both highest number of individual encounter histories data sets, the one based on pool- for comparison, but does not include the summer ing capture and observation and the of 2000. The second data set was based on capture other one using only capture data information only from the smaller part of the study from the smaller part of the study site, starting already in July 2000. Table 2.6 on site but spanning a longer period of page 29 gives detailed information on sample ses- time. sions. Although there were fewer individuals in this data set, the study period was longer. Infor- Pooled Capture mation on 142 animals could be used to estimate f m u f m u survival and recapture rates for the pooled data, 1999 4 0 1 6 2 0 and 65 animals for capture data respectively. 2001 2000 20 16 1 11 7 4 had the strongest cohort in both data sets (Ta- 2001 46 18 11 17 6 3 ble 4.9). These data sets were subjected to the 2002 21 4 0 8 1 0 same candidate sets of models. Coding of factors is illustrated in Tables 4.10 on the facing page and 4.11 on page 84. Age (a) was coded as ordinal variable into a “juvenile”, “adult”, and “senior” category. The categories were given the values 1 to 3, since they represent an order of growing individuals. For survival a linear and quadratic effect was supposed, to test if survival rates of juveniles and seniors were lower than those of adults. For recapture only a linear effect between juveniles and adults was tested. Age categories were delimited based on the life tables presented in Section 4.1.3 on page 74. Animals were classified as juvenile during the summer months of their first year, adult from their first September to their third summer, where they reach 2 years of age, and senior starting from the beginning of their third winter. Density (d) was given as estimated before (Chapter 3, p. 47) for pooled data. For

82 4. LIFE HISTORY 4.2. MODELLING SURVIVAL RATES

Table 4.10: Coding for drought (p), season (w), age (a), and density (d) for modelling survival and recapture rates of individuals using pooled information from capture and observation. Age is given in the categories for not yet born (nb), juvenile (j), adult (a), and senior (se) individuals of each cohort separately (a1999 2002). For more information on the capture − occasions see Table 2.6 on page 29.

Month of capture occasion 2000 2001 2002 Model notification 11 01 04 05 06 07 08 09 07 07 p 0 0 0 1 1 1 1 1 0 0 w w w w s s s s w a s s b b b a1999 a a a a a a a se se se a2000 a a a a a a a a a a a2001 nb nb nb j j j j a a a a2002 nb nb nb nb nb nb nb nb j j d 20.6 22.0 19.8 19.8 46.1 26.2 21.1 22.4 14.6 16.3 a Extra category for survival, since the interval between September and July includes the next summer. b For recapture rate, the behaviour of adult and senior individuals is supposed to be the same. capture data the MNKA (Minimal number known alive) within the initial part of the study area was calculated as an index for density. Season (w) was delimited based on the findings on seasonal development of age structure (Section 4.1.2, p. 69). The time from beginning of September to end of April was considered winter, while the time from the beginning of May to the end of August was considered summer. The interval from September 2001 to July 2002 was treated as an extra category for survival since the time interval comprised winter and half of the following summer. The drought year (p) was set to begin in summer 2001 and end in winter 2001/2002. The reference times for recapture and survival were treated differently. Recapture rate was estimated at the date of the sample session, while survival rate described the interval between this sample session and the following one. Covariates were transformed before modelling to fit within the interval of [0,1].

Canditate set of models Recapture and survival rates were analysed successively, using a similar pattern (Mod- els 1–8 and 9–22 in Table 4.12, p. 85). The most parsimonious model resulting from comparing the recapture models was used for comparing the candidate set of survival models. The fully parameterised model was always compared to a reduced model including a continuous covariate, and then without the covariate (Skalski et al., 1993). This was done for the two covariates age and density, and their combination (Model subsets a–c in Table 4.12). The resulting most parsimonious model from both the recapture and the survival model set was further reduced by excluding interactions, including only single factors, and ﬁnally holding recapture or survival constant for all individuals and all times of encounters. For the recapture models, the time factor was ﬁrst reduced to season only, then time was left out completely. Note that age always includes a component of time, so

83 4.2. MODELLING SURVIVAL RATES 4. LIFE HISTORY

Table 4.11: Coding for drought (p), season (w), age (a), and density (d) for modelling survival and recapture rates of individuals using capture data from the small part of the study site on which information is available for a longer period of time. For more information on the capture occasions see Table 2.6 on page 29. Age is given in the categories for not yet born (nb), juvenile (j), adult (a), and senior (se) individuals of each cohort separately (a1999 2002). − Month of capture occasion 2000 2001 2002 07 07 08 09 10 11 01 03 04 05 06 07 08 09 07 07 p 0 0 0 0 0 0 0 0 0 1 1 1 1 1 0 0 w s s s w w w w w w s s s s w a s s b b b a1999 a a a a a a a a a a a a a se se se a2000 j j j a a a a a a a a a a a a a a2001 nb nb nb nb nb nb nb nb nb j j j j a a a a2002 nb nb nb nb nb nb nb nb nb nb nb nb nb nb j j d 15 14 14 12 11 11 7 8 6 13 24 12 7 3 7 6 a Extra category for survival, since the interval between September and July includes the next summer. b For recapture rate, the behaviour of adult and senior individuals is supposed to be the same. time dependence is implicitly included in the models with age as explaining factor. This procedure was repeated for age, density, and the combination of age and density as covariates. For the survival candidate set of models, more factors were included. For each covariate the fully parameterised model was compared to reduced models with covariates, including the factors season and drought year. The fully parameterised model was successively reduced by ﬁrst replacing time of sample session with both factors, then with each of them alone, and then leaving both of them out. The candidate model set is listed in Table 4.12 on the next page.

Goodness of ﬁt for pooled and capture data

Although none of the χ2-tests could detect deviations from the assumptions of equal encounter and survival rates for the groups of individuals in the fully parameterised model, none of the groups had enough individuals to make the tests reliable. Bootstrap- ping data sets based on the parameters estimated by maximum likelihood resulted in deviances ranging from 101.77 to 174.15 for the data set based on pooling information from capture and observation. 93 of the simulated deviances were below the observed value of 159.4. The model thus can be accepted as not deviating significantly from the simulated data sets (p=0.07). Mean simulated deviance was 137.5, resulting in a value of 1.159 for cˆ for pooled data. Bootstrapping the most saturated model using capture data from the small part of the study site resulted in deviances ranging from 36.8 to 134.2. 95 of the simulated deviances were below the observed value of 112.9. The original data set was thus on the margin of being significantly different from the simulated data sets (p=0.05). Mean simulated deviance was 85.9, resulting in cˆ = 1.315.

84 4. LIFE HISTORY 4.2. MODELLING SURVIVAL RATES

Table 4.12: Candidate model sets for re- Recapture Survival capture (Models 1–8) and survival (Models (a) Modelling age dependence 9–22). “X” is a place holder for the most parsimonious model after running the re- 1 φ ρs time X φs c t ρ × × × capture model set (models 1–8). Recapture 2 φ ρs w a 9 φs w p a ρ × × × × × models tested sex (s), time (time), and sea- 3 φ ρs a 10 φs w a ρ × × × son (w) dependence. Survival models tested 4 φ ρs 11 φs p a ρ × × sex (s), cohort (c), time (time), season (w), 12 φs a ρ × and drought (p) dependence. Both model 13 φs ρ sets were run with the covariates (a) age (a), (b) Modelling density dependence (b) density (d), and (c) both age and density. 1 φ ρs time X φs c t ρ × × × 5 φ ρs w d 14 φs c w p d ρ × × × × × × 6 φ ρs d 15 φs c w d ρ × × × × 4 φ ρs 16 φs c p d ρ × × × 17 φs c d ρ × × 18 φs c ρ × (c) Modelling age and density dependence

1 φ ρs time X φs c t ρ × × × 7 φ ρs w a d 19 φs w p a d ρ × × × × × × × 8 φ ρs a d 20 φs w a d ρ × × × × × 4 φ ρs 21 φs p a d ρ × × × 22 φs a d ρ × × 13 φs ρ

Results Pooled data from capture and observation Table 4.13 on the following page summarises the results of comparing the models for the data set based on capture and observation. Taking only age as linear constraint for time in recapture rate (subset a), the best model included sex and age. It had 11 times more support than the next best model, which included only sex based on QAICc-weight. The influence of season and time was not supported by these models. None of the models using only density as linear constraint (subset b) supported the data better than the model using only sex as explaining factor. Although the combination of age and density (subset c) resulted in a model supporting the data four times more than the one using only sex, it did not surpass Model 3, which used age as single linear constraint. There was no difference between a model using only sex and a constant model, the difference in QAICc between Model 4 and Model 3d was less than two. Excluding the interaction between sex and age did not improve the support for the model, but using only age as explaining factor slightly improved the support for the data. There was no difference between the model using sex and age (3a) and the model using only age (3c), since the difference in QAICc was less than two. For further model comparisons, recapture rate was always constrained by age only. Table 4.14 on page 87 shows the modelling results of the candidate model set for survival. Using only age as linear constraint for cohort and time in survival rate (subset a), the best model included again sex and age (Model 12). Note that age included a linear and a quadratic term in this case as explained in the methods. It was 1.5 times better supported than the next best model, which included also the drought

85 4.2. MODELLING SURVIVAL RATES 4. LIFE HISTORY

Table 4.13: Models of the recapture candidate set for information based on pooled data from capture and observation. The candidate set is described in Table 4.12 on the preceding page. Values for QAICc weight of all models, including those of the survival candidate set shown in Table 4.14 on the next page add up to one, the model with the highest support having the highest score in QAICc weight. Models were selected in two steps, ﬁrst running all models of the candidate set (Models 1-8), and then optimising the most parsimonious model. The models with highest support within both steps are shown in bold letters. Model selection was based on cˆ = 1.159. Notation: sex (s), cohort (c), time (t), drought year (p for precipitation), season (w for winter), age (a as linear or quadratic ordinal covariate), and density (d). An underscore of denotes constant survival or recapture. K are the number of · estimable parameters.

QAICc Model QAICc ∆QAICc weight K Deviance (a) Modelling age dependence 1 φs c t ρs time 465.56 22.60 0.000 36 158.4 × × × 2 φs c t ρs w a 458.12 15.17 0.000 28 171.3 × × × × 3 a φs c t ρs a 451.00 8.05 0.011 25 171.5 × × × 4 φs c t ρs 455.43 12.47 0.001 26 173.5 × × (b) Modelling density dependence 1 φs c t ρs time 465.56 22.60 0.000 36 158.4 × × × 5 φs c t ρs w d 460.77 17.81 0.000 29 171.5 × × × × 6 φs c t ρs d 457.64 14.68 0.000 27 173.3 × × × 4 φs c t ρs 455.43 12.47 0.001 26 173.5 × × (c) Modelling age and density dependence 1 φs c t ρs time 465.56 22.60 0.000 36 158.4 × × × 7 φs c t ρs w a d 458.94 15.98 0.000 31 164.7 × × × × × 8 φs c t ρs a d 453.15 10.20 0.004 28 166.3 × × × × 4 φs c t ρs 455.43 12.47 0.001 26 173.5 × × Optimising the most parsimonious model 3 a φs c t ρs a 451.00 8.05 0.011 25 171.5 × × × 3 b φs c t ρs+a 457.62 14.66 0.000 27 173.3 × × 3 c φs c t ρa 450.62 7.66 0.013 23 175.9 × × 3 d φs c t ρ 455.40 12.44 0.001 25 175.9 × × · year as explaining factor. Using only density as explaining factor (subset b) produced generally better models than using only age. The best model was the one using sex, cohort and density (Model 17). It was 1.3 times better supported than the one using sex and age (Model 12), but both did not diﬀer by more than 2 in ∆QAICc. Since Model 12 used less parameter, it is preferred at this stage. But combining density and age (subset c) substantially increased the support for the data. The model using sex, age, and density (Model 22a) supported the data 13.4 times more than Model 17. Reducing this model did not increase the support for the data. The most parsimonious model after comparing all candidates thus included sex, age (with a linear and a quadratic component), and density as explaining factors for survival rates. A direct inﬂuence of season, drought year, cohort, or time of sample session was not supported by the data. It further included only age (with a linear com-

86 4. LIFE HISTORY 4.2. MODELLING SURVIVAL RATES

Table 4.14: Models of the survival candidate set for information based on pooled data from capture and observation. The candidate set is described in Table 4.12 on page 85. Values for QAICc weight of all models, including those of the survival candidate set shown in Table 4.13 on the facing page add up to one, the model with the highest support having the highest score in QAICc weight. Models were selected in two steps, ﬁrst running all models of the candidate set (Models 9-22), and then optimising the most parsimonious model. The models with highest support within both steps are shown in bold letters. Model selection based on cˆ = 1.159. Notation: sex (s), cohort (c), time (t), drought year (p for precipitation), season (w for winter), age (a as linear or quadratic ordinal covariate), and density (d). An underscore of denotes constant survival or recapture. K are the number of estimable · parameters.

QAICc Model QAICc ∆QAICc weight K Deviance (a) Modelling age dependence 3 c φs c t ρa 450.62 7.66 0.013 23 175.9 × × 9 φs w p a ρa 452.00 9.04 0.007 12 202.3 × × × 10 φs w a ρa 453.36 10.41 0.003 12 203.6 × × 11 φs p a ρa 449.74 6.79 0.020 10 204.4 × × 12 φs a ρa 448.73 5.77 0.034 7 209.7 × 13 φs ρa 484.37 41.42 0.000 4 251.6 (b) Modelling density dependence 3 c φs c t ρa 450.62 7.66 0.013 23 175.9 × × 14 φs c w p d ρa 449.50 6.54 0.023 19 184.1 × × × × 15 φs c w d ρa 449.42 6.46 0.024 18 186.3 × × × 16 φs c p d ρa 451.21 8.25 0.010 16 192.6 × × × 17 φs c d ρa 448.23 5.27 0.044 12 198.5 × × 18 φs c ρa 461.31 18.35 0.000 8 220.2 × (c) Modelling age and density dependence 3 c φs c t ρa 450.62 7.66 0.013 23 175.9 × × 19 φs w p a d ρa 452.11 9.15 0.006 16 193.5 × × × × 20 φs w a d ρa 453.33 10.38 0.003 16 194.7 × × × 21 φs p a d ρa 449.15 6.19 0.028 14 195.0 × × × 22 a φs a d ρa 442.96 0.00 0.609 10 197.6 × × 13 φs ρa 484.37 41.42 0.000 4 251.6 Optimising the most parsimonious model 22 a φs a d ρa 442.96 0.00 0.609 10 197.6 × × 22 b φs+a+d ρa 445.70 2.75 0.154 6 208.8 22 c φa+d ρa 457.21 14.25 0.000 4 224.5 22 d φa ρa 467.29 24.33 0.000 3 236.6 22 e φd ρa 531.18 88.22 0.000 2 302.5 22 f φs+d ρa 461.94 18.99 0.000 5 227.1 22 g φ ρa 496.41 53.45 0.000 2 267.8 · ponent coding for juveniles and adults) for recapture rates. Using this model, recapture rates were estimated to be 0.652 ( 0.025 SE) for juveniles and 0.779 ( 0.038 SE) for adults. Juveniles thus had a lowerrecapture rate than adults based on pooled data from observation and capture. Survival estimates are given for three diﬀerent densities,

87 4.2. MODELLING SURVIVAL RATES 4. LIFE HISTORY 1.0 0.8 0.6 0.4 Probability of Survival 0.2 0.0 Sex female male unsexed female male unsexed Age juvenile juvenile juvenile adult adult adult Density 15 21 50 15 21 50 15 21 50 15 21 50 15 21 50 15 21 50

Figure 4.11: Survival estimates with standard errors for the most parsimonious model based on pooling data from observation and capture November 2000 to July 2002. The model included the factors sex, age, and density with all interactions. Density is given as individuals per hectare. the minimum density of 14.6 individuals/ha, the median density of 21.4, and the maximum density of 49.8 individuals/ha (Figure 4.11). Not for every sex, age, and density combination an estimate could be calculated because of unreliable errors caused by small sample sizes. None of the senior age group could be estimated. Female juveniles had generally lower survival rates than female adults at the same densities. Survival rate decreased with increasing density in both age classes. Female juvenile survival nevertheless was high for the observed population, ranging from 0.557 ( 0.105 SE) to 0.794 ( 0.103 SE) depending on density. Male juveniles did not have consistently lower survival rates than adult males. Male juvenile survival decreased with density, similar to female juvenile survival, but it increased with density for adult males. Male juvenile survival was lower than female juvenile survival, ranging from 0.255 ( 0.158 SE) to 0.614 ( 0.268 SE). For the unsexed animals, a comparison is hampered bythe lack of data. Unsexed juveniles at maximum density had lower survival rates than unsexed adults at median density. Survival of unsexed juveniles did not diﬀer from the survival of male juveniles at maximum densities. Female adult survival ranged from 0.744 ( 0.133 SE) to 0.941 ( 0.023 SE), decreasing with increasing density. Female adult surviv al was consistently higher than male adult survival, although male adult survival increased with increasing density from 0.458 ( 0.382 SE) to 0.803 ( 0.052 SE) and the value at highest density could not be estimated. The error at the lowest density was relatively high for male adult survival and included most of the range estimated for median densities (Figure 4.11).

Capture data from July 2000 to July 2002 Table 4.15 on the facing page shows the results from comparing the models estimating recapture rates of the candidate set for capture data. None of these models had much support by the data. The

88 4. LIFE HISTORY 4.2. MODELLING SURVIVAL RATES

Table 4.15: Models of the recapture candidate set for information based only on capture data from the small part of the study site. The candidate set is described in Table 4.12 on page 85. Values for QAICc weight of all models, including those of the survival candidate set shown in Table 4.16 on the next page add up to one, the model with the highest support having the highest score in QAICc weight. Models were selected in two steps, ﬁrst running all models of the candidate set (Models 1-8), and then optimising the most parsimonious model. The models with highest support within both steps are shown in bold letters. Model selection was based on cˆ = 1.315. Notation: sex (s), cohort (c), time (t), drought year (p for precipitation), season (w for winter), age (a as linear or quadratic ordinal covariate), and density (d). An underscore of denotes constant survival or recapture. K are the number of · estimable parameters.

QAICc Model QAICc ∆QAICc weight K Deviance (a) Modelling age dependence 1 φs c t ρs time 236.98 55.28 0.00000 31 101.2 × × × 2 φs c t ρs w a 212.26 30.56 0.00000 21 110.3 × × × × 3 φs c t ρs a 208.10 26.39 0.00000 19 112.0 × × × 4 φs c t ρs 211.45 29.75 0.00000 19 115.4 × × (b) Modelling density dependence 1 φs c t ρs time 236.98 55.28 0.00000 31 101.2 × × × 5 φs c t ρs w d 215.55 33.84 0.00000 21 113.6 × × × × 6 φs c t ρs d 217.17 35.47 0.00000 21 115.2 × × × 4 φs c t ρs 211.45 29.75 0.00000 19 115.4 × × (c) Modelling age and density dependence 1 φs c t ρs time 236.98 55.28 0.00000 31 101.2 × × × 7 φs c t ρs w a d 213.51 31.80 0.00000 22 108.5 × × × × × 8 a φs c t ρs a d 201.41 19.70 0.00005 19 110.0 × × × × 4 φs c t ρs 211.45 29.75 0.00000 19 115.4 × × Optimizing the most parsimonious model 8 a φs c t ρs a d 201.41 19.70 0.00005 19 110.0 × × × × 8 b φs c t ρ(s+a+d)2 216.23 34.52 0.00000 22 111.2 × × 8 c φs c t ρa d 210.64 28.93 0.00000 19 114.6 × × × 8 d φs c t ρa 208.75 27.05 0.00000 18 115.5 × × 8 e φs c t ρd 209.55 27.84 0.00000 18 116.3 × × 8 f φs c t ρ 209.32 27.61 0.00000 18 116.1 × × · most parsimonious model from the first subset (Table 4.12, p. 85) was the one using sex and age as factors (Model 3). In subset (b) the model taking only sex as factor was best supported, although less than Model 3 from the first subset, the difference in QAICc being more than 2. In subset (c) the most parsimonious model was the one including both age and density additional to sex (Model 8). Reducing Model 8 did not increase model support. In contrast to this, the analysis based on data from capture and observation, selected Model 3 as the most parsimonious model at this stage (Table 4.13, p. 86). Table 4.16 on the next page shows the results from modelling the survival candidate set. In the first subset (a) the full model had highest support. The same was true when

89 4.2. MODELLING SURVIVAL RATES 4. LIFE HISTORY

Table 4.16: Models of the survival candidate set for information based only on capture data from the small part of the study site. The candidate set is described in Table 4.12 on page 85. Values for QAICc weight of all models, including those of the survival candidate set shown in Table 4.15 on the preceding page add up to one, the model with the highest support having the highest score in QAICc weight. Models were selected in two steps, ﬁrst running all models of the candidate set (Models 1-8), and then optimising the most parsimonious model. The models with the highest support within both steps are shown in bold letters. Model selection based on cˆ = 1.315. Notation: sex (s), cohort (c), time (t), drought year (p for precipitation), season (w for winter), age (a as linear or quadratic ordinal covariate), and density (d). An underscore of denotes constant survival or recapture. K are the number of · estimable parameters.

QAICc Model QAICc ∆QAICc weight K Deviance (a) Modelling age dependence 8 a φs c t ρs a d 201.41 19.70 0.00005 19 110.0 × × × × 9 φs w p a ρs a d 214.22 32.51 0.00000 15 129.3 × × × × × 10 φs w a ρs a d 220.77 39.06 0.00000 12 143.6 × × × × 11 φs p a ρs a d 218.15 36.45 0.00000 14 135.8 × × × × 12 φs a ρs a d 221.35 39.64 0.00000 11 146.6 × × × 13 φs ρs a d 221.59 39.89 0.00000 9 151.7 × × (b) Modelling density dependence 8 a φs c t ρs a d 201.41 19.70 0.00005 19 110.0 × × × × 14 φs c w p d ρs a d 205.74 24.03 0.00001 17 115.3 × × × × × × 15 φs c w d ρs a d 210.49 28.79 0.00000 19 114.4 × × × × × 16 φs c p d ρs a d 212.77 31.06 0.00000 19 116.7 × × × × × 17 φs c d ρs a d 208.85 27.14 0.00000 16 121.2 × × × × 18 φs c ρs a d 210.44 28.73 0.00000 13 130.7 × × × (c) Modelling age and density dependence 8 a φs c t ρs a d 201.41 19.70 0.00005 19 110.0 × × × × 19 a φs w p a d ρs a d 181.71 0.00 0.99990 4 123.2 × × × × × × 20 φs w a d ρs a d 202.33 20.62 0.00003 12 125.1 × × × × × 21 φs p a d ρs a d 216.31 34.61 0.00000 16 128.7 × × × × × 22 φs a d ρs a d 217.81 36.11 0.00000 14 135.5 × × × × 13 φs ρs a d 221.59 39.89 0.00000 9 151.7 × × Optimizing the most parsimonious model 19 a φs w p a d ρs a d 181.71 0.00 0.99990 4 123.2 × × × × × × 22 b φs+w+p+d ρs a d 215.26 33.55 0.00000 12 138.1 × × 22 c φ(s+w+p+d)2 ρs a d 230.69 48.98 0.00000 20 131.7 × × constraining only by density (subset b). When age and density were both added as covariates, the model using the factors sex, season, and drought year additional to the covariates (Model 19) had much higher support than all the other models. At the same time the model could only estimate four parameters. It could not be further reduced without losing support for the data. The most parsimonious model thus included sex, age, and density in the recapture rates and sex, season, drought year, age, and density in the survival rates. These are

90 4. LIFE HISTORY 4.2. MODELLING SURVIVAL RATES

more factors in than the previous analysis, which used more animals over a shorter period of time. At the same time, the goodness of ﬁt test for this data set was worse than the one for the pooled data and in the most parsimonious model only few parameters could be estimated. Recapture rates estimated by the model decreased with density for juvenile females and increased with density for adult females. The same was true for male individuals. But for most of the estimates there were too few data to be reliable. For survival rates even less parameters could be estimated. Survival rates of adult females were higher than of juvenile females, although diﬀerent periods of time were compared: female survival in the winter before the drought for adults was higher than female survival in the summer of drought for juveniles.

Discussion The climate factors drought and season were considered irrelevant for explaining survival rates when using the data based on pooling information of capture and observation (Table 4.14, p. 87). At the same time, the factors drought and season were included in the most parsimonious model when using the data based on information of capture sessions alone (Table 4.16, p. 90). However, the results from the second models are dubious. The goodness of fit test showed significant deviance from the simulated models. Additionally, the overwhelming support of Model 19 is caused by the inability to explain more than four parameters. For this reason, the present study accepts the results derived from the first model based on information from capture and observation stating that the climate factors drought and season are considered irrelevant for explaining survival rates. Effects of past climatic conditions can result in different survival of cohorts coming from different years (Bjørnstad et al., 2004; Stenseth et al., 2002). Similar to the present effects of climate, represented by th drought and by seasonal differences, effects from past climatic conditions were not supported by the capture-recapture analysis, since cohort affiliation did not have a significant effect when explaining survival rates. Although the numbers of the animals of the the 2000-cohort were consistently higher than those of the other cohorts, this difference was not significant when analysing life tables, either (Section 4.1.3, p. 74). The effect may have been missed due to small sample size, though. Although the influence of season was not supported directly in the analysis, it was supported indirectly, since age had significant effects on survival rates. Age however was defined using seasons. Thus, the effects of season and age are coupled for the Mongolian pika. Juveniles show lower survival rates (Figure 4.11, p. 88) than adults and juveniles are only present in the summer season. Adults on the other hand were present in winter and summer seasons and here seasons could not explain differences in survival rates. It may even pay off to distinguish between younger and older juveniles within the summer season, since the life tables indicate increasing survival with age (Table 4.5, p. 75). Contrary to the expectations, juveniles showed lower recapture rates than adults. Observations during field work were the other way round, since juveniles seemed easier to capture, at least within one capture session. Having lower recapture rates between capture sessions for juveniles may be misleading. Low recapture may have resulted from the short period of time that an individual was supposed to be juvenile, since the same individual is assorted to another group for the analysis when it turns adult

91 4.2. MODELLING SURVIVAL RATES 4. LIFE HISTORY

(Table 4.10, p. 83). Adult female survival was consistently higher than adult male survival at low and median densities. However, while adult female survival decreased with density, it increased with density in males. Male survival rates at median densities were estimated to be higher than female survival rates at high densities (Figure 4.11, p. 88). Higher male survival was also detected when analysing life tables (Table 4.6, p. 76). Since highest density was reached in summer 2001, during the reproductive period, these differences in survival rates can be explained by the higher reproductive effort of females as opposed to males. Male pikas have been shown to invest little effort in rearing the young (Smith and Gao, 1991). But at the same time the higher survival rate of males may also be an artefact caused by the weight limit used to group individuals into cohorts. This weight limit is mainly based on growth data from females, and males exhibited a different growth pattern (Section 4.1.1, p. 58). Excluding drought from the factors influencing survival seems to be in contradiction to the apparent differences in density between the drought year 2001 and the following year. Even if the hay caches from the preceeding year could explain high survival during the summer of drought, they should have been used up by winter. For modelling survival rates, this winter was coded as an extra category within the factor coding season, since it included major parts of the following summer (Table 4.10, p. 83). But neither the drought factor nor that of season could explain the survival rates in the present analysis. One of the methodological reasons why an influence of the drought year on survival might have been missed by the analysis is sample size: Most of the individuals used for the analysis were captured in the summer of drought. There might have been simply too few individuals in the other groups to detect significant differences. This already was the case when using χ2 tests for analysing consistent recapture and survival rates within the groups, as described in the methods of this section. Another possible methodological reason for missing the influence of drought is the “juvenile” category described above. Judging from the life tables, survival rates increased within the group of juveniles with age (Table 4.5, p. 75). Juvenile survival could be only compared for the drought summer of 2001 and the following summer, but in 2002 there were only two sample sessions, both in the beginning of July. Again, the small sample size could be a reason for not detecting reduced survival in the drought year. However, since survival is not the only factor explaining population densities, the analysis may also be reliable and may not have been influenced by the drought year. Instead, reproduction may have been reduced in the year following the drought year. Section 4.5 on page 101 discusses the implications of accepting high adult survival and density dependence for the life history strategy and the population dynamics of the Mongolian pika.

4.2.3 Survival of early and late litter Within the concept of two distinct suits of life history traits for pikas higher survival of early litter was linked to the K-type of life history strategy (see Section 2.2, p. 23 and Smith, 1988). In the following analysis, encounter histories of juveniles born in the same summer are tested for the eﬀect of their date of birth within the season on survival. Recapture rates were tested for the inﬂuence of sex, date of birth, and age,

92 4. LIFE HISTORY 4.2. MODELLING SURVIVAL RATES

Table 4.17: Candidate model sets for re- (a) Recapture rates (b) Survival rates capture (Models 1–6) and survival (Models 7–11) using only individuals from early or 1 φ ρs b t X φs b t ρ × × × × late litter in summer 2001. “X” is a place 2 φ ρs b a 7 φs b a d ρ × × × × × holder for the most parsimonious model af- 3 φ ρb a 8 φs b ρ × × ter running the recapture model set (models 4 φ ρs b 9 φs a d ρ × × × 1–6). Recapture models tested sex (s), time 5 φ ρb 10 φs ρ of birth (b), capture occasion (t), and age 6 φ ρa 11 φb ρ (a) dependence. Survival models additionally tested density (d) dependence. while survival rates were tested for the inﬂuence of sex, date of birth, age, and density.

Methods Data from juveniles was based on pooled information from observation and capture occasions. Only juveniles from the summer of 2001 were included into the analysis. Classification to a group of early or late born individuals was based on weight at first capture, as described in Section 4.1.3, p. 74. 17 juveniles were considered to be early born, 11 late born. Encounter histories started with the beginning of reproduction in May 2001, and ended in July 2002. There were 7 encounter occasions used for this analysis (see Table 2.6, p. 29 for more information on encounter occasions). Anal- ysis was done using maximum likelihood techniques and capture-recapture analysis implemented in program MARK (White and Burnham, 1999), as described in Sec- tion 4.2.1 on page 79. The candidate set of models is shown in Table 4.17. The fully parameterised model showed acceptable goodness of fit. Bootstrapped deviances from the most saturated model ranged from 3.474 to 35.065. Observed deviance of the model was 23.726. 79 out of 100 simulated deviances were lower than the observed deviance (p=0.21). Although the software program RELEASE did not detect violation of the hypotheses, none of the χ2-tests had sufficient numbers to be reliable. For this reason, cˆ was calculated from the bootstrapped and observed deviances. Mean bootstrapped deviance was 18.4, resulting in a correction factor for overdispersion (cˆ) of 1.289 (for methodological background see Section 4.2.1, p. 79).

Results All the models manipulating recapture rate were better supported by the data than the full model (Table 4.18 on the next page). None of them diﬀered by more than 2 in ∆QAICc, they can thus be considered equivalent. The model with the least parameters was Model 6, using the linear constraint of age to explain recapture rate. Neither survival model was better supported than the full model explaining survival with sex, birth, and time. Reducing this model to an additive one did not improve explanation. Model 6 was thus accepted as best model to explain survival and recapture rate for the animals born in 2001 from early and late litter. Only four of the estimates could be given with reliable error margins. All of them were from juvenile females still in the year 2001 (Figure 4.12). Although the estimates for the individuals from early litter were both higher than those of the individuals

93 4.2. MODELLING SURVIVAL RATES 4. LIFE HISTORY

Table 4.18: Model likelihoods for modelling survival and recapture for individuals from early and late litter in the summer of 2001. The candidate set of models is described in Table 4.17 on the preceding page. Values for QAICc weight of all models add up to one, the model with the highest support having the highest score in QAICc weight. Models were selected in two steps, ﬁrst running all models of the recapture candidate set (Models 1-11), and then optimising the most parsimonious model. The models with highest support is shown in bold letters. Model selection based on Cˆ = 1.289. Notation: sex (s), time of birth (b), time of capture occasion (t), age (a as linear or quadratic ordinal covariate), and density (d). An underscore of denotes constant survival or recapture. K are the number of estimable · parameters.

QAICc Model QAICc ∆QAICc weight K Deviance (a) Recapture rate 1 φs b t ρs b t 90.14 6.92 0.007 11 18.4 × × × × 2 φs b t ρs b a 84.56 1.35 0.119 8 21.8 × × × × 3 φs b t ρb a 84.56 1.35 0.119 8 21.8 × × × 4 φs b t ρs b 83.22 0.00 0.234 7 23.2 × × × 5 φs b t ρb 83.22 0.00 0.234 7 23.2 × × 6 φs b t ρa 83.91 0.69 0.166 7 23.9 × × (b) Survival rate 6 φs b t ρa 83.91 0.69 0.166 7 23.9 × × 7 φs b a d ρa 103.85 20.64 0.000 12 28.8 × × × 8 φs b ρa 85.88 2.66 0.062 4 33.5 × 9 φs a d ρa 90.43 7.21 0.006 7 30.5 × × 10 φs ρa 87.76 4.55 0.024 3 37.7 11 φb ρa 94.95 11.73 0.001 3 44.9 Optimizing the most parsimonious model 6 b φs+b+t ρa 87.53 4.31 0.027 7 27.6 from late litter, the error margins were wide. The results discussed in Section 4.5 on page 101.

4.2.4 Summary Capture-recapture analysis including the estimation of recapture probabilities presents a more powerful tool to analyse presence absence data than mere enumeration summarised in life tables. Using Cormack-Jolly-Seber models (White and Burnham, 1999) implemented in modelling environments that use maximum likelihood theory to estimate survival and reencounter parameters the effect of age, cohort, affiliation to early or late litter, sex, season, drought conditions, and density on survival rates were estimated. Neither climatic factor could explain the differences in survival rates: the models using season or drought conditions were less supported by the data. Past climatic conditions indicated by different survival of cohorts were not supported either. The most parsimonious model included age, sex, and density in modelling survival rates, and age in modelling recapture rates.

94 4. LIFE HISTORY 4.2. MODELLING SURVIVAL RATES 1.0 0.8 0.6 0.4

Probability of survival May − Jun 0.2 Jun − Jul Figure 4.12: Survival rates for female ju- Jul − Aug veniles born early or late in the summer.

0.0 Rates could be compared for three intervals early late between sample sessions, May to June, June to July, and July to August 2001. Estimates Early or late litter are given with standard errors.

For all groups survival decreased with density, the only exception being adult males. Adult male survival could be considered to remain constant for varying densities. Fe- male survival was consistently higher than male survival, except for adult female and male survival at maximum densities. At maximum densities male survival was higher. Juvenile survival was consistently lower than adult survival. Additionally, survival of juveniles from early litter was higher than of juveniles from late litter. Female juvenile survival ranged from 0.557 ( 0.105 SE) to 0.794 ( 0.103 SE) depending on density, male juvenile survival ranged from 0.255 ( 0.158 SE) to 0.614 ( 0.268 SE). Female adult survival ranged from 0.744 ( 0.133 SE) to 0.941 ( 0.023 SE), while male adult survival ranged from 0.458 ( 0.382 SE) to 0.803 ( 0.052 SE). Survival rates were standardised to time intervals of 30 days.

95 4.3. REPRODUCTION 4. LIFE HISTORY

4.3 Reproduction

The reproductive potential of a species is crucial for assessing possible scenarios of its population dynamics (Krebs, 1985). Of the two life history strategies described for pika species, one invests more energy in reproduction than the other. These species reach a reproductive output one order of magnitude greater than species of the other group (Section 2.2, p. 23, Table 2.4, p. 24, and Smith, 1988). This section describes the reproductive activity of the Mongolian pika.

Methods Data on reproductive activity was collected while handling individuals (Section 2.4.2, p. 28). Information from all capture occasions was used (Table 2.6, p. 29). The status of reproductive activity for females was assessed by the appearance of their mamilla, namely if these were merely visible, already swollen, or if the hair around the papilla was glued together, indicating that they were lactating. For males, the testicles were tested. They were either merely tangible, or visible, or swollen. Individuals with no signs of reproductive activity were not included in the analysis. For each month from July 2000 to July 2002 the sum of individuals with signs of reproductive activity was calculated. Information on reproductive output was collected by counting juveniles during the capture and observation sessions in 2001 (Table 2.6 on page 29), including additional captures on selected burrows in the middle of June 2001. Litter size was determined by the number of juveniles observed on one burrow. Individual juveniles were assigned to the burrow they were observed on up to a weight of approximately 80 g. If juveniles were observed on more than one burrow, their number was divided by the number of burrows. Individuals found dead were included in the analysis. Additionally, the number of females per burrow with litter was counted.

Results Reproductive activity None of the animals captured as juvenile showed signs for reproduction in the summer of birth. In April 2001 females showed ﬁrst signs of reproductive activity in form of visible mammilla (Figure 4.13 on the facing page). Lactating females could be detected in May and June 2001. One year before, in July 2000 there was one animal with visible mammilla. Note that capture took place on the smaller part of the study site in July and August 2000. Lactating females were also detected in July 2002. Males showed signs of reproductive activity over a longer period of time than females. The ﬁrst visible testicles were observed in March 2001, the last in July 2001, which is one month earlier and later than signs on females. In 2000 testicles were still visible in August (Figure 4.13).

Litter Of the 29 burrows on the trapping site (Section 2.4.1, p. 26), 17 showed juveniles (59 %). One of the burrows had 2 litters. Minimum observed litter size was 1, maximum 7, and median 3 young per litter. On the whole study site in summer 2001 53 juveniles were observed, 4 of them were found dead. This results in an average of 1.7 living juveniles per burrow, and 2.9 living juvenile per burrow with litter. In one

96 4. LIFE HISTORY 4.3. REPRODUCTION

10 Females visible 8 swollen

6 lactating 4 2 0 Sep Jan May Sep Jan May

10 Males tangible 8 visible

6 swollen Number of individuals 4 2 0 Sep Jan May Sep Jan May 2001 2002 Dates of capture occasions

Figure 4.13: Reproductive state of captured animals during the research period from July 2000 to July 2002. Reproductive state refers to the appearance of mamilla or testicles. Months with capture occasions are labelled with “ ”. ◦ case a female could be observed with two litters separated in time. Although all other observed pikas did not use more than one burrow, this female used 2 burrows. It also happened to be nearest to the observation point (3 m), so that everything could be seen in great detail. This female produced the highest number of juveniles. On these two burrows 13 juveniles were observed.

Discussion The observed population showed a discrete dynamic in sexual activity and reproductive output. Both were restricted to the summer months. Sexual activity was easier to detect for males than for females, since females showed only signs of pregnancy and lactation, but no signs of fertility. Sexual activity started in March and ended in August. This period of sexual activity is consistent with the weight structure of the observed population (Section 4.1.1, p. 58). Note the exceptional heavy female in April 2001, which was probably pregnant. In August the weight range of April was reached again, implying that all juveniles had reached the adult weight range. The length of the reproductive season in the Gobi Altai may be too short to allow young of the year to reproduce. Young of the year could not be observed reproducing in the drought year 2001. Data was too sparse for the other years to make inferences on reproduction of young of the year. Since reproduction in the year of birth has been observed for both, Ochotona pallasi pallasi (Shubin) and Ochotona pallasi pricei (Krylova, 1974 quoted in Smith, 1988), this may be possible in the observed population, too, but only under especially favourable circumstances. Since survival was not

97 4.3. REPRODUCTION 4. LIFE HISTORY influenced by the drought conditions in the summer of drought (Section 4.2, p. 79) and population density in this summer was considerably higher than in the year after the drought (Chapter 3, p. 47), it is not expected that reproductive output was influenced substantially by the drought conditions directly. Additionally, female young of the year might be limited by the scarcity of territories, even if they were capable of reproduction as far as their physical condition was concerned. There were 40 g animals which were ambiguous to assort to one burrow, indicating that the observed individuals were not tolerated within the home range of their mother for a long time. Another possible reason for not detecting reproduction by young of the year is that the research area is located at the limit of their distribution. On the other hand, the Mongolian pika is a ubiquitous animal in the National Park (Reading, 1996), indicating that the habitat quality is sufficient to populate this area. Relying on the observed data there were 17 females and 9 males present in the summer of 2001 (see the life table in Table 4.6, p. 76), resulting in a sex-ratio of 1.9 females to males. At the same time, population density on the study site was estimated to be 19.8 animals in April 2001, before reproduction started. Using the above sex-ratio, these were 12.9 females and 6.9 males. Since juveniles were observed on 17 burrows of the trapping site, every female should have reproduced at least once. A litter size of 3 young is in accordance with the observation of Millar (1973) in Ochotona princeps. This species looses embryos when the count is greater than 3 and these losses are not attributed to nutritional stress but rather the capacity of the uterus. On the other hand, other authors observed more young per female in Ochotona p. pricei (up to 12; Lazarev, 1968). Observation effort probably played a role in detecting the juvenile pikas. Highest reproductive output (2 litter with a total of 13 young) was observed with the individual nearest to the observation point. Monthly observation seems too long a period of time to detect all reproductive output. Many females have not been captured during pregnancy and lactation, since signs have been seen on only a few individuals. Juve- nile survival seemed to be low at a very early stage, since those juveniles which could be followed through time in capture-recapture analysis had high survival rates (Fig- ure 4.11, p. 88). They probably died before they could be captured. During June and July, many weak and dead juveniles have been found on the study site. But even if the number of young per litter has been underestimated by this study, the number of litter per female was still low compared to other pika species adopting the burrowing type of life history strategies (see Table 2.4, p. 24). See Section 4.5 on page 101 for further discussion on life history traits.

98 4. LIFE HISTORY 4.4. POPULATION DYNAMICS

4.4 Population dynamics

Possible scenarios for population dynamics of the Mongolian pika are crucial for assessing their “pest” status (Sections 1.4, p. 4, and 2.2, p. 23). The following section integrates the ﬁndings on survival and reproduction of the previous sections resulting in a simple model for population dynamics for the Mongolian pika.

Methods A model of population dynamics using discrete time steps in a Leslie- Matrix (Krebs, 1985) approach is used to extrapolate population dynamics. This approach calculates the number of individuals in distinct age classes for each time step using fertility and survival rates. Calculations were performed using the software R (R Development Core Team, 2004). Fertility rates result from reproduction and survival of the young weaned. Maximal reproductive output observed in the studied population was used, while female/male ratio was expected to be even. Survival rates for all age classes were taken from the parameter estimates of the most parsimonious model resulting from capture-recapture analysis (see Section 4.2.2, p. 81). Male survival increased with density, but the error margin was high for low densities (Figure 4.11, p. 88). For this simulation, adult male survival was assumed to be constant and high. Time steps were the same as used for constructing life tables (Section 4.1.3, p. 74) and modelling survival rates, a summer consisting of 4 months and a winter consisting of 8 months. Life time and age classes of pikas correspond to those used in modelling survival and those proposed from interpreting life tables: a juvenile category spanning the ﬁrst summer, and adult category spanning the ﬁrst winter to the third summer (1 to 2.5 years), and a senior category for the third (and last) winter. Survival rates of female (sf ) and male (sm) individuals were dependent on age (juvenile: j and adult: a) and density (d), according to the modelling results (Sec- tion 4.2.2, p. 81). The following equations for density dependence are based on a 2nd order polynomial regression on the parameter estimates.

s = 0.87 0.40 (d/100) 0.44 (d/100)2 (4.4) fj − · − · s = 0.80 1.16 (d/100) + 0.17 (d/100)2 (4.5) mj − · · s = 0.96 + 0.04 (d/100) 0.93 (d/100)2 (4.6) fa · − · sma = 0.802 (4.7)

Transition rates are given in units of 30 days. Summer survival rates would therefore 4 8 be s , and winter survival rates s . The Leslie matrix requires sx, the survival rate from one age group x to age group x + 1, and fx, the number of oﬀspring (bx) born in one time interval per female alive aged x to x + 1 and surviving (sx) to enter the ﬁrst age group at the next time interval. Since a maximum of 13 juveniles per female was observed in 2001 (Section 4.3 on page 96), 6.5 female and 6.5 male juvenile per adult or senior female were assumed, leading to the following equation for female juveniles (ff ): f = b s = 6.5 s4 . (4.8) f f fj · fj In this equation there is still a density dependence for female juveniles, since their

survival (sfj , Equation 4.4) decreases with density. Similarly, male fm is calculated as

f = b s = 6.5 s4 , (4.9) m f mj · mj 99 4.4. POPULATION DYNAMICS 4. LIFE HISTORY

using Equation 4.4 for male juvenile survival. Survival rates for adult females and males were given in equations 4.6 and 4.7 on the page before. These assumptions lead to the following Leslie matrices for the number of females and males in the population. The number of females in each age class after one step of time is determined by the number of females present, their survival rates, and the number of female juveniles being born and surviving this time step:

0 f 0 f 0 f fa0.5 f f f fa0.5 st 0 0 0 0 0 fa1.0 fa fa1.0    t    fa1.5 0 sfa 0 0 0 0 fa1.5 = t (4.10)  fa   0 0 s 0 0 0   fa   2.0   fa   2.0  f 0 0 0 st 0 0 f  a2.5   fa   a2.5     t     fs   0 0 0 0 s 0   fs   3.0   fa   3.0        In contrast, the number of males in each age class after one step in time depends on both, the number of males and the number of females present. While the number of juvenile males is determined by the number of males born and surviving the ﬁrst step in time, the number of adult males is determined by the survival rates of adult males (Equation 4.7, p. 99):

ma0.5 0 fm 0 fm 0 fm fa0.5

ma1.0 0 0 0 0 0 0 fa1.0  m   0 0 0 0 0 0   f  a1.5 = a1.5 +  ma   0 0 0 0 0 0   fa   2.0     2.0   ma   0 0 0 0 0 0   fa   2.5     2.5   m   0 0 0 0 0 0   f   s3.0     s3.0        0 0 0 0 0 0 ma0.5 t 0.802 0 0 0 0 0 ma1.0  t    0 0.802 0 0 0 0 ma1.5 t (4.11)  0 0 0.802 0 0 0   ma2.0   t     0 0 0 0.802 0 0   ma     2.5   0 0 0 0 0.802t 0   m     s3.0  To explore the populationdynamics resulting from the above formulations, the matrix system was iterated for 100 years for four diﬀerent starting values within a range of possible values based on the observations made during the study period. Since population densities before and after the reproductive period, in April and August 2001 were 19.8 and 21.1 animals respectively, a range of possible numbers of pikas on the trapping site was assumed to be within one and 30 animals for one sex. Out of this range, four starting points for the iteration were chosen, namely one female and one male, 30 females and 10 males, 30 females and 30 males, and ﬁnally 10 females and 30 males.

Results and Discussion All chosen starting points lead to densities oscillating between 38 animals (26 females and 11 males) per ha at the end of the summer period, and 9 animals (7 females and 2 males) at the end of the winter periods (Figure 4.14). This corresponds well with the values observed during the study period, where densities varied between 14.6 and 46.1 animals (Chapter 3, p. 47). Nonetheless, the low summer densities in the summer after the drought cannot be explained with this model. In summer 2002 densities ranged between 14.6 and 16.3 animals. The following section further discusses the implications of such a density dynamic for the life history traits of Mongolian pikas.

100 4. LIFE HISTORY 4.5. DISCUSSING LIFE HISTORY STRATEGY 30 20 Number of males 10 0

0 10 20 30 40 50

Number of females

Figure 4.14: Trajectories for starting points featuring numbers of female and male individuals. Numbers were chosen to represent the range of combinations which could be observed in real situations (own judgement). Equations 4.10 and 4.11 on the preceding page describe the population dynamics from the starting points. The model was run to simulate 200 years with summer and winter seasons. 4.5 Discussing the life history strategy of the Mon- golian Pika

Smith et al. (1990) and Smith (1988) propose a trade-oﬀ for all pika species between high investment in present survival and high investment in present reproduction (Sec- tion 2.2, p. 23). Similar to the concept of r/K-selection (Section 1.3, p. 3) they link this trade-oﬀ to the quality of the habitats the species evolved in. Relatively stable habitats enable stable population densities near the carrying capacity, leading to higher investment in present survival, while in relatively unstable habitats survival is unpredictable, leading to higher investment into present reproduction during the evolution

Figure 4.15: Two pikas ﬁghting on the trapping site in July 2002. The individual facing the reader is a two- year-old female, while the individual with the back to the reader is a one-year-old male.

101 4.5. DISCUSSING LIFE HISTORY STRATEGY 4. LIFE HISTORY

Table 4.19: Life history traits of burrowing and non-burrowing pikas in general as summarised by Smith (1988) and for Ochotona pallasi pricei from literature together with the ﬁndings of this study.

Burrowing Ochotona pallasi pricei Life history traits no yes Literature This study

Reproduction litters 1–3 2–5 3a 1–2 litter size 1–7 1–13 1–12b 1–7 in summer of birth no yes yesa no Survival juveniles low high ? low late litter lower higher ? lower adults high low ? high overwinter high low ? high density dependent yes no ? yes Age structure old age 6 2 4c 2 % yearlings low high highc high Density dynamics density (/ha) 2.6–30 0–380 ? 14.6–46.1 density variation low high highd low density regulation territories climate ? territories drought

a Krylova (1974) quoted in Smith (1988) b Lazarev (1968) quoted in Smith (1988) c Krylova (1973) d Zevegmid (1975)

of the species. Based on the available literature they show that pika species, which dig their own burrows and live in steppe environments exhibit life history traits of an r-strategist, while non-burrowing pikas exhibit life history traits of a K-strategist (Section 2.2, p. 23). The findings of the present study confirm the proposed distinction between two opposing life history strategies in pikas, since the Mongolian pika exhibits life history traits in accordance with the K-type of life history strategy of non-burrowing pikas. At the same time, the findings do not confirm the link between a burrowing life-style and the life history strategy, since the Mongolian pika is a burrowing species. This section discusses why the Mongolian pika can be considered to belong to the group of non-burrowing pikas as proposed by Smith (1988), while the link between a burrowing lifestyle and the quality of the habitat of the Mongolian pika will be discussed following the analysis of the impact of the species on the ecosystem (Chapter 6, p. 133).

How a burrowing pika can be a non-burrowing pika Table 4.19 lists the life history traits studied in this thesis grouped for questions concerning reproduction, survival, age structure, and density dynamics. Life history traits

102 4. LIFE HISTORY 4.5. DISCUSSING LIFE HISTORY STRATEGY

related to reproduction are the number of litter, size of litter, and if animals reproduce already in the same summer, in which they were born. The number of litter was lower than expected from a typical burrowing species (Table 4.19), and young of the year were not observed to reproduce. Although the species has the potential to reach high fecundity rates with litter sizes up to 7 young, most of the females had far less offspring (Section 4.3, p. 96). Reproductive data from this study do not agree with data from the literature on the same species, though. In other studies, up to three litters are observed, as well as reproduction of young of the year (Krylova, 1973 quoted in Smith, 1988). Addi- tionally, most of the data in this study were collected under drought conditions. The females may have been energy-limited under these conditions, reducing their ability to produce a second or third litter. It was shown earlier in this study that the reproductive season was delayed in the summer of drought (Figure 4.4, p. 64 and Section 4.1.1, p. 58). This shortens the reproductive period and reduces the chances for young of the year to reproduce in their summer of birth. Energy limitation may be indicated by comparatively low weight gain or a lower juvenile ratio than in the other years. Pikas gained less weight in the drought year than in the year before since median weight was lower in August and September 2001 than in 2000 (Table 4.1, p. 61). Similarly, juvenile ratio was lower in 2001 than in the other years, but this was not significant (Table 4.4, p. 71). On the other hand, the data collected outside the summer of drought confirms the findings on comparatively low reproduction for the Mongolian pika. The more favourable year 2000 did not show a markedly different weight structure indicating reproduction by young of the year in a year less limited by food. Like in the year of the drought there were no juveniles weighing less than 100 g in August or September (Figure 4.1, p. 60), although the animals reached higher median weights, indicating good physical conditions in that year (Table 4.1, p. 61). Additionally, pikas are less energy-limited under drought conditions than other species. Goats and sheep studied in the research area also gained weight during the drought summer, although not enough to build up resources for the winter (Retzer et al., in press). At the end of the summer, many herders left the area and tried to sell some of their weak animals. Livestock numbers were thus dramatically reduced by the end of the summer. In contrast, pika numbers were not reduced compared to the other years. Energy limitation is thus less crucial for pikas compared to livestock in a year of drought. This is partly due to the hay stores collected by pikas, but also to their capacity to graze more thoroughly than livestock (Farnsworth et al., 2002; Retzer, 2004). For this reason the present study concludes that the relatively low investment in reproduction observed in the Mongolian pika is representative for the population, even though major parts of the information were collected in a drought year. Survival rates also conform with the K-type of life history pattern of non-burrowing species (Table 4.19, p. 102). Non-burrowing species show lower survival of juveniles, and high adult survival. Overwinter survival is high in non-burrowing species, while it is low in burrowing species (Section 2.2, p. 23 and Smith, 1988). The low temperatures in winter do not pose a problem to pikas in general as pikas are cold-adapted species. Like other small mammal species they are capable of nonshivering thermogenesis, lib- erating chemical energy mainly generated by brown adipose tissue: Wang and Wang (1996) show that Ochotona curzoniae do not reduce nonshivering thermogenesis in summer and that they don’t reduce body weight in winter, a strategy employed to

103 4.5. DISCUSSING LIFE HISTORY STRATEGY 4. LIFE HISTORY

save energy in other small mammals. Low overwinter survival rates of the burrowing species Ochotona curzoniae have been related to high and prolonged snow cover during winter, preventing the animals from foraging (Wang and Wang, 1996). In the observed population of Ochotona pallasi pricei survival of juveniles was lower than adult survival and overwinter survival was high (Section 4.2.2, p. 81). During winter there was snow covering the ground and pika burrows. Pikas digged tunnels in the snow and runways on the snow surface were visible (own observation). Mongolian pikas obviously are able to reach their hay caches or standing crop in spite of snow cover. Higher survival rates for early born juveniles (Figure 4.12, p. 95) also conform with the life history traits of non-burrowing species. The drought conditions are another explanation for low survival of late born juveniles, though. Probably both, the drought conditions and the need to defend territories limited the energy available for pikas to raise young. Additionally, this life history trait may be variable even in burrowing species, since Wang and Smith (1988) report higher survival rates for late born juveniles in Ochotona curzoniae. In the present study the effect of density on survival was stronger for juveniles than for adults. At the same time, a study on the age structure of the Mongolian pika (Krylova, 1973) shows that in pellets of birds of prey jaws of younger individuals are more prominent than those of older ones. This indicates that safe sites from predation, and thus territories with burrows, are crucial for survival of juveniles of this species. While non-burrowing pikas show strong density dependence in survival, burrowing pikas do not (Table 4.19, p. 102). In burrowing species, density increases during the summer months, with juveniles of subsequent litters staying within the home-range of their parents. Survival then decreases during winter (Smith, 1988). In contrast, survival rates in non-burrowing species are attributed to aggressive territorial behaviour and not harsh winter conditions. Juveniles are tolerated for a shorter time by their mother than in non-burrowing species. Additionally they partition resources in time, being active during different parts of the day to avoid contact (Smith, 1988). Although the observed population of Ochotona pallasi pricei proved to be density-dependent (Section 4.2.2, p. 81), high density was again coupled with the summer of drought, since highest densities were reached in the summer of 2001. However, density changed within the drought summer and still explained survival rates, so the effect of density dependence can be shown over and above the effect of the drought. At the same time, aggressive territorial behaviour is documented for Ochotona pallasi pricei (Proskurina et al., 1985) and has been observed on the study site (Figure 4.15, Monkhzul, 2005). Additional to the low juvenile survival rates documented in this study, many dead juveniles have been found on the study site during late summer. There were some very weak juveniles in the traps. Parasite load was higher during the summer, when population density was also high (Section 2.5, p. 40). This indicates that late summer and autumn are the crucial periods with high mortality for the Mongolian pika. This means that mortality is independent of the availability of forage, since not only forage in form of standing crop and hay stores is still available in late summer and autumn but also adult individuals reach highest body weights during this time (Table 4.1, p. 61). Life span and age structure are the only life history traits corresponding more with the burrowing-type of pikas. None of the observed individuals reached 3 years of age, consequently, the proportion of juveniles and yearlings was high in the population (Ta- ble 4.4, p. 71 and Section 4.1.3, p. 74). Although Shubin also reports high proportions of juveniles in populations of Ochotona pallasi pallasi, the data come from the summer

104 4. LIFE HISTORY 4.5. DISCUSSING LIFE HISTORY STRATEGY

months. He reports an increase from 30 % at the end of May to 90 % at the end of July. These numbers correspond with the proportions observed in the present study for the year 2000 (Table 4.4, p. 71). But if survival is coupled to the possession of territories, a proportion of 90 % juveniles in July may be reduced markedly until September. At the same time, there remained considerable doubt about the classification of the individuals to cohorts in the year 2000, since trapping started as late as July and most of the individuals classified as “juvenile” were captured much later in the year (Section 4.1.2, p. 69). Nonetheless, even if a percentage of juveniles as high as 80 % may be reached only during the reproductive months, a maximal age of two years still corresponds more with the burrowing species (Table 4.19, p. 102). It contradicts the findings of Krylova (1973) though. She reports high percentages of two to four year-old individuals, namely 20 % of the population. Also implications for weight development differ, since Krylova (1973) assumes that individuals continue to grow in their second summer, while the present study has shown that individuals reach their adult weight within their summer of birth, even under drought conditions (Section 4.1.1, p. 58). Longterm studies are needed to detect if these differences are caused by intraspecific variation or if the two populations belong to different species. Last but not least density dynamics of the observed population conformed with that of non-burrowing species (Table 4.19, p. 102). Although the drought conditions, and thus environmental variability, obviously reduced population densities with the time lag of one year (Section 3, p. 47), they did not effect the survival of individuals (Sec- tion 4.2, p. 79). The observed individuals must have cached sufficient plant biomass in their burrows to mitigate the effect of the drought. Caching hay is a common strategy for non-hibernating small mammals to survive the winter in steppe environments (Formozov, 1968). For the Mongolian pika it additionally can serve as resource during adverse conditions even in the following summer. It thus has the function of a key-resource comparable to areas with extra water supply for larger herbivores (Illius and Connor, 1999). Such key-resources make it possible to maintain higher densities, even in years without enough forage due to unfavourable conditions for plant growth. In the present study, survival rates were not effected by the drought conditions (Section 4.2, p. 79) leading to predictions of stable but oscillating population densities between 50 individuals per ha in summer and 20 in winter (Section 4.4, p. 99). These predictions are well in accordance with the observed range of population densities of 14.6–46.1 individuals per ha, except for the low population densities in the summer of 2002. Thus, population densities predicted by density-dependent survival alone do not explain the observed variation in population densities. At the same time, the observed variation in population densities is still nearer to the variation observed for the non-burrowing pika species (Table 4.19). Although none of the tests about different cohort survival was significant, the cohort of 2000 still had higher proportions than the other cohorts within each of the years it was observed (Table 4.4, p. 71). Having non-significant tests may simply be due to low sample sizes (Section 4.2, p. 79). Holst et al. (1999) have shown that rabbits with high social rank improve their reproductive success and their lifetime independent of present reproductive effort. This was explained by lower psychosocial stress responses in high ranking individuals. Maybe the cohort of 2000 gained higher social rank and could thus survive more easily, due to the favourable conditions in 2000. Observations must be continued over a longer period of time to disentangle the effect of density and

105 4.5. DISCUSSING LIFE HISTORY STRATEGY 4. LIFE HISTORY resource availability on survival and fecundity rates of the Mongolian pika. There are several other factors controlling population densities which could not be tested by the present analysis, including reproductive rates and predation. Lower population densities could have been caused by lower reproductive rates in the summer following the drought year rather than by lower adult survival. The influence of predation could not be studied within this thesis. Although pikas obviously form an important source of prey at the study site and evolved different mechanisms to escape predation (Lutton, 1975), the influence of predation on population dynamics is not expected to be great. Predation has been shown to control lower trophic levels only when ecosystems are productive enough (Begon et al., 1996; Oksanen and Oksanen, 1990), which is not the case in a dry steppe environment. Observations from the study site are in accordance with this prediction, since many dead and weak pikas have been found during late summer. This indicates that other factors than predation were important for low survival during that period of time. Density-dependence itself is a much debated topic. While some authors adopt a pragmatic approach, arguing that density dependence is given when the rate of increase in a population decreases with density (Berryman et al., 2002), others feel that there is no density dependent regulation at all since every density is ultimately regulated by the available resources (White, 2004). For the Mongolian pika such a resource may be the availability of territories. Non-burrowing pikas show highly aggressive behaviour towards individuals within their home ranges (Smith et al., 1990). Within the species Ochotona pallasi, the subspecies pricei shows aggressive relationships 10 times more frequently than the subspecies pallasi. When individuals trespass onto the territories of neighbours, they may be killed by bites inflicted by the occu- pant (Formozov and Proskurina, 1980 quoted in Smith et al., 1990). Proskurina et al. (1985) attribute the higher rate of aggression and the smaller overlap of territories found in Ochotona pallasi pricei to environmental qualities, in this case habitats poor in food and resting places. This is in accordance with the comparatively low population densities in the year before the drought as observed in the present study. There obviously is an upper limit for the population density, independent of climatic conditions, probably controlled by the availability of territories. Large enough territories then enable individuals to collect enough plant material to survive not only the winter, but also the following summer if conditions are adverse.

Life history strategy and habitat quality Finding a burrow-dwelling pika in a variable environment to exhibit life history traits conforming with the K-type of life history strategies contradicts the link between habitat quality and life history strategy as suggested by (Smith, 1988). The group of species in pikas exhibiting life history traits similar to an r-strategist is named “burrowing” pikas, or “steppe-dwelling” pikas in Smith (1988); Smith et al. (1990). The link between constructing burrow and r-strategy is rather cursory, saying that the proximity of forage supports directing more energy into reproduction for burrowing species. The underlying reason for the evolution of the two life history patterns is actually seen in the predictability of the habitat, similar to the r/K-concept (Section 1.3, p. 3). While steppe habitats require the construction of burrows for pikas because of the absence of safe sites provided by structures like rock-piles or trees, they require an r-type of life history strategy because of the unpredictability in climatic conditions. Now, ﬁnding

106 4. LIFE HISTORY 4.5. DISCUSSING LIFE HISTORY STRATEGY

Ochotona pallasi pricei to be a K-strategist in a steppe environment might imply that (a) the species did not evolve in the steppe environment it is found in today, or (b) that the studied habitat is not as unpredictable as might be expected, or (c) the trade-oﬀ between present survival and present reproduction is not linked to the unpredictability of the habitat. These three hypothesis will be discussed in Chapter 6 on page 133.

107 4.6. SUMMARY 4. LIFE HISTORY

4.6 Summary

The present study is the first one to observe individually marked Mongolian pikas (Ochotona pallasi pricei). It includes information on individual pikas collected during three summers and one winter from the year 2000 to 2002. One of the summers was a heavy drought. Reproductive activity was observed from March to August, while signs of female pregnancy were evident from April to July. First juveniles appeared in May, juveniles grew to an adult weight between 180-200 g within two months. Although it may be physiologically possible for individuals of the species to reproduce already in their summer of birth, this was not observed in the present study. All females gave birth, with a median of 3 and a maximum of 13 juveniles per female. The longest observed life-span was 2.3 years. Using capture-recapture analysis with maximum likelihood techniques an effect of age, density, and sex on survival rates could be shown, while there was no effect of cohort affiliation nor of the climatic factors season and drought. However a classification of a summer and a winter season was useful to delimit age groups. Generally, adults showed higher survival rates than juveniles, females showed higher survival rates than males, and in all groups survival declined with density. Although the drought conditions did not influence survival rates directly, density was reduced in the year after the drought. This means that the animals could delay the effects of the drought by one year most probably by relying on their stocks of cached hay from the year before. Using the parameter estimates for survival in a model of population dynamics based on a system of Leslie-matrices, there was a stable but oscillating attractor for starting points ranging between one female and one male, and 30 females and 30 males on one hectare. After 100 years of simulating population dynamics, densities oscillated between 38 animals (26 females and 11 males) per ha at the end of the summer period, and 9 animals (7 females and 2 males) at the end of the winter period. Comparing reproductive effort, factors influencing survival rates, and density dynamics of the Mongolian pika with other species of the genus shows that this species exhibits traits of the group of non-burrowing pikas, which are closer to a K-type of life history strategy than the group of burrowing pikas.

108 Chapter 5

Ecosystem impact

In arid non-equilibrium systems, herbivores are not expected to exert much influence on ecosystem dynamics. They are in contrast seen as driven factors, dependent on climatic conditions, primarily precipitation (Wesche and Retzer, 2005). Nonethe- less, small mammals as well as larger herbivores are often regarded as threats to pasture conditions from the perspective of pastoral land use (Hilbig and Opp, 2005; Illius and Connor, 1999), or as threats to ecosystem functioning from the perspective of nature conservation (Madhusudan, 2004; Reading et al., 2001). At the same time, small mammal burrowing activity can be beneficial for pasture quality (Kinlaw, 1999). The ecosystem impact of an organism is often dependent on its density (Begon et al., 1996). Because of their territoriality, long-term pika densities can be indicated by the density of their burrows (Chapter 4, p. 57). At the same time, climatic factors as well as biotic interactions can influence burrow densities. Altitudinal gradients present gradients of precipitation and productivity. Additionally, small mammals can increase in density in areas already overgrazed by livestock (Samjaa et al., 2000; Zhong et al., 1985). This chapter focuses on plant performance and biomass removal by pika and livestock on burrow and steppe habitat (Section 5.1, p. 109), and on the influence of altitude and livestock densities on pika burrow densities (Section 5.2, p. 124).

5.1 Plant biomass production and biomass removal by herbivores

Although both, grazing and burrowing activity, effect ecosystem qualities, burrowing activity is usually given more importance in arid systems (Jones et al., 1997; Kinlaw, 1999; Lavrenko et al., 1993; Whitford and Kay, 1999). Nutrient levels are often enhanced on small mammal burrows, as is plant productivity and soil disturbance (Wesche et al., submitted; Whitford and Kay, 1999). Higher plant productivity can be beneficial or detrimental from the perspective of livestock. There are examples of small mammal burrows on which fodder plants flourish (Grinnell, 1923; Holtmeier, 1999), while others host weeds (Holtmeier, 1999; Samjaa et al., 2000). Thus, the productivity and the usage of burrow habitat should be taken into account when assessing the impact of pikas on the ecosystem. Given drought conditions, there is severe competition between livestock and pikas on steppe habitat (Retzer, 2005). The following analysis studies productivity and

109 5.1. PRODUCTION AND BIOMASS REMOVAL 5. ECOSYSTEM IMPACT

biomass removal by livestock and pikas on burrow and steppe habitat under varying precipitation.

Methods The study was performed in the surroundings of the research camp (Section 2.1 on page 13) on upper pediments at 2300 m asl with a southern exposure and low inclination (3-6 ). The vegetation of the pediments consists of species from desert steppes (Stipa gobica, Allium polyrrhizum) and mountain steppes (Agropyron cristatum, Stipa krylovii, Wesche et al., in press). Most of the pika burrows on the slightly sloping pediments are dominated by Agropyron cristatum (Wesche et al., submitted). They cover 7-12 % of the surface in the study region. Mongolian pikas (Ochotona pallasi pricei) are the dominant small mammals (Section 2.4, p. 26). Their densities were comparatively high during the drought year 2001, while they were relatively low during the study period for the exclosure experiment presented here (Table 5.1). Both, wild large herbivores and livestock graze in the surroundings of the research camp, but livestock are more abundant (Section 2.1, p. 13 and Retzer et al., in press). Under drought conditions as in the year 2001 there is strong competition between livestock and pikas, the latter being the better competitors because they can graze more thoroughly (Retzer, submitted). The exclosure experiment consisted of two sets of Table 5.1: Number of pikas 4 plots for vegetation sampling, one set of plots on each of captured on the trapping 10 pika burrows and one set of plots adjacent to them in grid (100 100 m2, Chapter 3, × steppe habitat. Maximal distance between selected bur- p. 47) from the year 2001 to rows was 1000 m, minimal distance was 10 m. Plots were 2003. Data from 2003 are circular with an area of 0.27 m2. Each of the 4 plots had from T. Monkhzul. a diﬀerent treatment, with access for only one of the her- Date Pikas bivore groups (pikas or large herbivores), none, or both 23./24.6.2001 62 of the herbivore groups. This was achieved by choosing 27./28.7.2001 31 a meshed wire for the exclosures and covering the whole plot with it for exclosure of pikas as well as herbivores. 29.6.-2.7.2002 22 Plots with wire only at the sides, but not above, could 7./8.7.2002 17 be accessed by larger herbivores, while plots covered with 3.-8.7.2003 20 meshed wire but cut open at the sides could be accessed 19./20.7.2003 7 by pikas only. The fourth plot was not fenced at all and could thus be accessed by both herbivore groups. In each plot we estimated cover for each plant species. Cover values were estimated visually, estimates were given directly as percentage, results proved to be fairly close among diﬀerent researchers after a short period of practice. Plots were randomly placed on the steppe but not on burrows, where entrances of pikas and open earth were avoided. Cover values were thus representative of steppe habitat, but not of burrow habitat. Burrow vegetation covers from representative samples taking an area of 3 3 m2 had a median of 5.48 % (Wesche, unpublished data). Median cover on steppe ×plots was 3.76 % (Table 5.2). In the two summers data on 23 plant species were collected. These were categorised in the following groups for harvesting and analysis: (1) Stipa spec. (2) Agropyron cristatum (3) Allium spec. (4) remaining. Table 5.2 on the facing page gives median values for vegetation cover of all plots, medians for species group cover and median cover for the species contained in the group of remaining plants, as estimated in July 2002, when the plots were established. These groups were chosen since Agropyron cristatum

110 5. ECOSYSTEM IMPACT 5.1. PRODUCTION AND BIOMASS REMOVAL

Table 5.2: Median cover values on the sample plots before the start of the exclosure experiment in July 2002. Sample size was four burrow and four steppe plots on and adjacent to each of 10 burrows. The Allium spec. group consisted of A. polyrrhizum and A. prostratum, the Stipa spec. group consisted of S. gobica and S. krylovii.

Species burrow steppe Species burrow steppe Vegetation Carex stenophylla 0.10 0.07 14.97 3.76 Chenopodium spec. 0.01 0.04 Cleistogenes songorica 0.03 0.04 Groups Convolvulus ammanii 0.12 0.09 Agropyron cristatum 10.50 1.30 Dontostemon spec. 0.05 – Allium spec. 0.98 0.70 Heteropappus hispidus – 0.15 Stipa spec. 0.20 0.50 Kochia prostrata – 0.55 Remaining 0.60 0.76 Lepidium densiﬂorum 0.04 – Oxytropis spec. 0.02 0.05 Remaining plants Pedicularis spec. 0.60 0.10 Arenaria meyeri 0.10 0.10 Ptilotrichum canescens 0.10 0.05 Artemisia frigida 0.50 0.30 Salsola colina 0.02 0.01 Astragalus spec. 0.35 0.25 Scorzonera ikonnikovii 0.10 0.10 Axyris spec. 0.01 – not identiﬁed 0.01 – Bupleurum bicaule 0.02 0.01

dominates mountain steppes and montane desert steppes of the Gobi-Gurvan-Saikhan Park (Wesche et al., in press), it accounted for up to 40 % of the total vegetation on steppe plots and 86 % of the vegetation cover on burrow plots. Allium spec. cover were highest after A. cristatum. The plants come from the onion-family with strong smelling secondary compounds. Allium spec. is so common in the Gobi desert that Mongolians say they can recognise meat from this region by its taste. Stipa spec., although not covering as much as the other two groups, are typical species for steppe environments. All three groups are important fodder plants for livestock (Retzer, 2004). In 2002 above-ground plant biomass was harvested in July before fencing the plots, and again in August after Table 5.3: Precipitation one month of treatment. Plant biomass was harvested in the summers of 2002 for the whole plot without separating plant groups. In and 2003. Data from Wesche and Retzer (2005). July and August 2003 plant groups were separated as described above when harvested from the same plots. Rain (mm) Thus, time intervals between harvest events were different. While there was about one month between the August 2002 0.0 July 2003 31.8 July and August harvests in 2002 and 2003, there were August 2003 62.2 11 months between August 2002 and July 2003. Since vegetation growth does not begin before May at the study site, a time interval of 2.5 months was assumed as growing period preceeding the July harvest in 2003. Plant biomass was oven-dried at 105 ¢ in Halle, Germany. 160 out of 768 samples from 2003 could not be oven-dried due to moulding. These samples were estimated based on median losses obtained from the available samples separately for plant groups. Median drying coeﬃcients were 0.86 for Agropyron cristatum, 0.55 for Allium spec., 0.77 for Stipa spec., and 0.81 for the remaining plants. While there was no precipitation up to

111 5.1. PRODUCTION AND BIOMASS REMOVAL 5. ECOSYSTEM IMPACT

Figure 5.1: Hierarchical design Time for a) the data set comparing the two summers and b) the data Pika grazing set comparing species groups in Habitat summer 2003. The highest hierarchical level were the burrow sites, within which there were Livestock grazing burrow and steppe sites (habitat). Within the habitat sites there were plots grazed by live- a) stock and plots not grazed by livestock. Within these there Species group were again plots grazed by pikas and not grazed by pikas. The Pika grazing last hierarchical level was har- Habitat Time vest time within one sample plots for both data sets, but in the case of comparing species groups Livestock grazing they represent another hierarchical level between pika grazing and harvest time. b)

August 2002, the precipitation increased for the harvest times in July and August 2003 (Table 5.3). Samples were analysed using analysis of variance for a split plot design. Two different models were calculated, one for total plot biomass for August 2002, and July and August 2003, the other for July and August 2003 only, but including species groups in the analysis (Figure 5.1). In the first model, burrow site was the largest experimental unit and formed the blocks within which the treatments took place. Each burrow site was split into burrow and steppe habitats. Each of these plots was again split into two plots with and without livestock access. In each of these treatments, one plot was accessible by pikas. Time was treated as the last split plot within each of the experimental units (Figure 5.1). In order to account for the differences in vegetation cover on the sample plots, initial cover in July 2002 at the beginning of the experiment was added as a covariate. Burrow site was treated as random block effect and added to the model in the error term. Time was treated as fixed factor, since there was information on precipitation available for the time intervals and we were interested in the effects produced by the different harvesting times. Contrasts were calculated to compare the effect of sampling time. First, the harvest without precipitation (Au- gust 2002) was contrasted to the harvests with precipitation (July and August 2003). After this, July 2003 was contrasted to August 2003. The second model had a very similar structure, but harvest time only had two levels, so that contrasts were not calculated. Species groups were treated as split plot before the level of harvest time. Since there were more than two species groups, contrasts were used to separate first Allium spec. from the other groups. In a second contrast, Agropyron cristatum was separated from Stipa spec. and the group of remaining plants, and in a last step, those two groups were contrasted with each other (Figure 5.1). Box-cox transformations were used to achieve a better error distribution before calculating the models. They resulted in a λ of 0.1625 for the first model, and a λ

112 5. ECOSYSTEM IMPACT 5.1. PRODUCTION AND BIOMASS REMOVAL of -0.1161 for the second model. The following transformations were thus used to transform values of biomass (b):

bλ 1 b = − (5.1) t λ Estimates of standing above-ground biomass were averaged using a linear model including only the significant terms of the split plot analysis. To compare the effects of burrow and steppe habitat, grazing type, species group, and harvest time, estimates were corrected for the influence of initial vegetation cover by standardising on a cover of 8 %.

Estimating plant production and biomass removal Estimates for above ground plant biomass on burrow and steppe plots with median vegetation cover were used to assess the biomass production and biomass removal by pikas on one hectare mountain steppe including pika burrows. The term “biomass removal by pikas” is preferred to “pika grazing” here, since it includes both, grazing and harvesting. Since burrows occupied between 7 and 12 % of the surface in our study region (Wesche and Retzer, 2005), the mean of 9.5 % was used as value for the area covered by burrows on a hectare. Median vegetation cover was 3.76 % on steppe plots (Table 5.2, p. 111) and 5.48 % on burrow plots (data from 3 3 m2 relev´es, Wesche, unpublished data). While above-ground plant biomass production× was estimated from the biomass harvested on the ungrazed plots, biomass removal by pikas was estimated by the diﬀerence between the ungrazed plots and the plots grazed by pikas. Production and removal could be related to a period of time by using the time intervals between the harvest events. There was one month of growth for the harvests in August 2003 and 2003, while it was approximated as 2.5 months in July 2002. This corresponds to daily intervals of 30 and 75 days respectively.

Results In both models, initial cover, pika grazing, sampling time, and habitat were significant factors explaining variation in standing crop, while livestock grazing missed significance marginally. In the second model, plant species groups also contributed significant effects (Tables 5.4 and 5.5). Pooling all other factors in the first model, initial cover could not explain much of the variation on the level of burrow sites, but discriminating between burrow and steppe habitat, initial cover explained most of the variation (Table 5.4). On the level of livestock grazing treatments, neither of the factors contributed significantly to explain variance, although the interaction between habitat and livestock grazing nearly reached significance. Pooling sampling times, pika grazing showed a significant main effect. Above ground plant biomass on pika-grazed plots was lower than on plots not grazed by pikas. Figure 5.2 on page 115 shows estimates for dried biomass on a standardised initial vegetation cover of 8 % based on the linear model featuring only the significant effects of the split plot ANOVA. Letters “A” and “B” represent the main effect of pika grazing and pika exclusion respectively. Sampling time, and thus the amount of precipitation, had not only significant main effects but also significant interactions with habitat and pika grazing. Above-ground

113 5.1. PRODUCTION AND BIOMASS REMOVAL 5. ECOSYSTEM IMPACT

Table 5.4: Analysis of variance table on biomass harvested on exclosure plots in 2002 and 2003. Significant effects are printed in bold letters. Significance codes: 0 < ‘***’ < 0.001 < ‘**’ < 0.01 < ‘*’ < 0.05 < ‘.’ < 0.1 < ‘’ < 1. variable Df Sum Sq Mean Sq F value Pr-(> F ) Sign. Error: site arcsin(√cover) 1 0.047 0.047 0.03 0.864 Residuals 8 12.050 1.506 Error: site:habitat arcsin(√cover) 1 65.482 65.482 98.50 < 0.001 *** habitat 1 1.473 1.473 2.22 0.175 Residuals 8 5.319 0.665 Error: site:habitat:lvst arcsin(√cover) 1 0.220 0.220 1.08 0.312 lvst 1 0.306 0.306 1.50 0.237 habitat:lvst 1 0.869 0.869 4.27 0.054 . Residuals 17 3.455 0.203 Error: site:habitat:lvst:pika arcsin(√cover) 1 0.004 0.004 0.02 0.902 pika 1 4.599 4.599 18.37 < 0.001 *** habitat:pika 1 0.284 0.284 1.13 0.294 lvst:pika 1 0.541 0.541 2.16 0.150 habitat:lvst:pika 1 0.082 0.082 0.33 0.570 Residuals 35 8.762 0.250 Error: site:habitat:lvst:pika:time Aug 02 1 240.459 240.459 993.20 < 0.001 *** Jul 03 1 2.956 2.956 12.21 0.001 *** habitat:Aug 02 1 0.905 0.905 3.74 0.055 . habitat:Jul 03 1 1.149 1.149 4.75 0.031 * lvst:Aug 02 1 0.075 0.075 0.31 0.578 lvst:Jul 03 1 0.177 0.177 0.73 0.394 pika:Aug 02 1 0.890 0.890 3.68 0.057 . pika:Jul 03 1 2.374 2.374 9.81 0.002 ** habitat:lvst:Aug 02 1 0.003 0.003 0.01 0.919 habitat:lvst:Jul 03 1 0.052 0.052 0.22 0.643 habitat:pika:Aug 02 1 0.689 0.689 2.85 0.094 . habitat:pika:Jul 03 1 0.348 0.348 1.44 0.232 lvst:pika:Aug 02 1 0.032 0.032 0.13 0.716 lvst:pika:Jul 03 1 0.126 0.126 0.52 0.471 habitat:lvst:pika:Aug 02 1 0.050 0.050 0.20 0.652 habitat:lvst:pika:Jul 03 1 0.080 0.080 0.33 0.567 Residuals 144 34.863 0.242

plant biomass on the sample plots increased from August 2002 to August 2003, similar to the increasing precipitation (Table 5.3, p. 111) and regardless of pika grazing. The interannual diﬀerence between the harvest in August 2002 and the two harvests in 2003 explained most of the variation. Letters “E” to “G” in Figure 5.2 on the facing

114 5. ECOSYSTEM IMPACT 5.1. PRODUCTION AND BIOMASS REMOVAL

A B A B A B A B A B A B C C D D

E E E E F F F F G G G G 50 ]

2 H I m g 40 no pika grazing pika grazing 30 20 10 Oven dried biomass [ 0 August 2002 July 2003 August 2003 burrow steppe burrow steppe burrow steppe

Figure 5.2: Averaged estimates for standing crop on a standardised initial cover of 8% based on a linear model taking into account only the significant effects of the analysis of variance. The letters indicate significant main effects and interactions: plots grazed by pikas (A), plots not grazed by pikas (B), the effect of pika grazing at different harvest times, namely July 2003 (C) and August 2003 (D), the main effect of harvest time in August 2002 (E), July 2003 (F), and August 2003 (G), and the different effects of habitat type at different harvest times, in July 2003 (H) and in August 2003 (I). page indicate the main effect of harvest time on the estimated above-ground biomass, pooling habitat and grazing treatment. Burrow plots produced more plant biomass than steppe plots when comparing July and August 2003 (interaction between habitat and July 2003 in Table 5.4, p. 114). This interaction is indicated by letters “H” and “I” in Figure 5.2, showing that most of the higher plant biomass produced up to August 2003 originates from burrow plots, not from steppe plots. Additionally, pika biomass removal was more evident in July 2003 than in August 2003 (interaction between pika grazing and July 2003 in Table 5.4). This effect is indicated by the letters “C” and “D” in Figure 5.2. The second model, using data only from the summer of 2003, found the same effects (pika and time main effect, interaction between pika grazing and time, and interaction between habitat and time). Species groups could explain additional variation (Ta- bles 5.5 and 5.6). When pooling all other factors, Allium spec. showed highest above- ground biomass production, followed by Agropyron cristatum, the group of remaining plants, and finally the Stipa spec.-group. In contrast to the other species, Allium spec. showed a consistently higher above-ground biomass production on plots not grazed by pikas (interaction between pika grazing and Allium spec.), while A. cristatum was removed more than the other plants. A. cristatum showed higher biomass production than Allium spec. on ungrazed plots (interaction between pika grazing and Allium spec.). The effect of preferential grazing of A. cristatum was even more evident on pika burrows (interaction between pika grazing, habitat, and A. cristatum). The group of remaining plants showed higher biomass production on burrow plots grazed by pikas, while pikas reduced this plant group on steppe plots (interaction pika grazing, habitat, remaining plants). See Table 5.6 on page 120 for further significant effects and Fig- ure 5.3 on page 119 for estimates of biomass production for the factors explaining the variation in the data and corrected for effects of initial vegetation covers.

115 5.1. PRODUCTION AND BIOMASS REMOVAL 5. ECOSYSTEM IMPACT

Table 5.5: Analysis of variance table on biomass harvested on exclosure plots in July and August 2003, including species groups as explaining variable. Significant effects are printed in bold letters. Significance codes: 0 < ‘***’ < 0.001 < ‘**’ < 0.01 < ‘*’ < 0.05 < ‘.’ < 0.1 < ‘’ < 1.

variable Df Sum Sq Mean Sq F value Pr-(> F ) Sign. Error: site arcsin(√cover) 1 0.005 0.005 0.01 0.940 Residuals 8 6.140 0.768 Error: site:habitat arcsin(√cover) 1 9.239 9.239 14.15 0.006 ** habitat 1 0.782 0.782 1.20 0.306 Residuals 8 5.225 0.653 Error: site:habitat:lvst arcsin(√cover) 1 0.038 0.038 0.16 0.690 lvst 1 0.182 0.182 0.79 0.387 habitat:lvst 1 0.468 0.468 2.03 0.173 Residuals 17 3.928 0.231 Error: site:habitat:lvst:pika arcsin(√cover) 1 0.010 0.010 0.04 0.845 pika 1 3.069 3.069 11.91 0.001 ** habitat:pika 1 0.011 0.011 0.04 0.838 lvst:pika 1 0.554 0.554 2.15 0.151 habitat:lvst:pika 1 0.003 0.003 0.01 0.909 Residuals 35 9.016 0.258 Error: site:habitat:lvst:pika:group arcsin(√cover) 1 110.694 110.694 316.14 < 0.001 *** Allium 1 49.851 49.851 142.37 < 0.001 *** Agrop. 1 2.742 2.742 7.83 0.006 ** Remain. 1 11.274 11.274 32.20 < 0.001 *** habitat:Allium 1 0.405 0.405 1.16 0.284 habitat:Agrop. 1 1.035 1.035 2.96 0.087 . habitat:Remain. 1 0.153 0.153 0.44 0.509 lvst:Allium 1 0.381 0.381 1.09 0.298 lvst:Agrop. 1 0.003 0.003 0.01 0.924 lvst:Remain. 1 0.030 0.030 0.09 0.770 pika:Allium 1 4.059 4.059 11.59 0.001 *** pika:Agrop. 1 4.287 4.287 12.24 0.001 *** pika:Remain. 1 0.667 0.667 1.90 0.169 habitat:lvst:Allium 1 0.479 0.479 1.37 0.243 Continued on next page

116 5. ECOSYSTEM IMPACT 5.1. PRODUCTION AND BIOMASS REMOVAL

Continued from last page variable Df Sum Sq Mean Sq F value Pr-(> F ) Sign. habitat:lvst:Agrop. 1 0.031 0.031 0.09 0.766 habitat:lvst:Remain. 1 0.003 0.003 0.01 0.923 habitat:pika:Allium 1 0.101 0.101 0.29 0.592 habitat:pika:Agrop. 1 1.378 1.378 3.93 0.049 * habitat:pika:Remain. 1 2.099 2.099 5.99 0.015 * lvst:pika:Allium 1 0.960 0.960 2.74 0.099 . lvst:pika:Agrop. 1 0.018 0.018 0.05 0.821 lvst:pika:Remain. 1 0.054 0.054 0.15 0.695 habitat:lvst:pika:Allium 1 0.482 0.482 1.38 0.242 habitat:lvst:pika:Agrop. 1 0.006 0.006 0.02 0.892 habitat:lvst:pika:Remain. 1 0.486 0.486 1.39 0.240 Residuals 215 75.281 0.350 Error: site:habitat:lvst:pika:group:time time 1 1.550 1.550 12.61 < 0.001 *** habitat:time 1 1.231 1.231 10.01 0.002 ** lvst:time 1 0.008 0.008 0.06 0.802 pika:time 1 2.233 2.233 18.17 < 0.001 *** Allium:time 1 0.018 0.018 0.15 0.699 Agrop.:time 1 0.095 0.095 0.77 0.381 Remain.:time 1 1.889 1.889 15.37 < 0.001 *** habitat:lvst:time 1 0.020 0.020 0.16 0.690 habitat:pika:time 1 0.488 0.488 3.97 0.047 * lvst:pika:time 1 0.021 0.021 0.17 0.679 habitat:Allium:time 1 0.119 0.119 0.97 0.326 habitat:Agrop.:time 1 0.001 0.001 0.01 0.933 habitat:Remain.:time 1 0.355 0.355 2.89 0.090 . lvst:Allium:time 1 0.019 0.019 0.15 0.696 lvst:Agrop.:time 1 <0.001 <0.001 <0.01 0.970 lvst:Remain.:time 1 0.007 0.007 0.05 0.816 pika:Allium:time 1 0.381 0.381 3.10 0.079 . pika:Agrop.:time 1 0.131 0.131 1.07 0.302 pika:Remain.:time 1 0.541 0.541 4.40 0.037 * habitat:lvst:pika:time 1 0.279 0.279 2.27 0.133 habitat:lvst:Allium:time 1 0.154 0.154 1.25 0.264 habitat:lvst:Agrop.:time 1 0.030 0.030 0.24 0.623 habitat:lvst:Remain.:time 1 0.341 0.341 2.77 0.097 . habitat:pika:Allium:time 1 0.496 0.496 4.03 0.046 * habitat:pika:Agrop.:time 1 0.049 0.049 0.40 0.527 habitat:pika:Remain.:time 1 <0.001 <0.001 <0.01 0.963 lvst:pika:Allium:time 1 0.113 0.113 0.92 0.339 lvst:pika:Agrop.:time 1 0.061 0.061 0.50 0.482 Continued on next page

117 5.1. PRODUCTION AND BIOMASS REMOVAL 5. ECOSYSTEM IMPACT

Continued from last page variable Df Sum Sq Mean Sq F value Pr-(> F ) Sign. lvst:pika:Remain.:time 1 <0.001 <0.001 <0.01 0.978 habitat:lvst:pika:Allium:time 1 0.006 0.006 0.05 0.820 habitat:lvst:pika:Agrop.:time 1 0.113 0.113 0.92 0.339 habitat:lvst:pika:Remain.:time 1 0.002 0.002 0.01 0.908 Residuals 288 35.399 0.123

Table 5.7, p. 121 lists estimates for daily above-ground plant biomass production scaled up to a hypothetical hectare of a mountain steppe. Burrows occupy 0.095 ha on this plot. Production was highest in August 2003 and lowest in August 2002 (main effect of harvest time in Table 5.4). Pika biomass removal was also lowest in August 2002 (0.12 kg d-1), but was highest in July 2003 (1.27 kg d-1; interaction between pika grazing and the contrast between July and August 2003 in Table 5.4). Although pika numbers were relatively constant in 2002 and 2003 (Table 5.1, p. 110), they removed about 10 times more above-ground plant biomass in July 2003 than in August 2002. Above-ground plant biomass production on this hypothetical hectare added up to 1.11 kg d-1 in August 2002, from which 0.24 kg d-1 (22 %) were produced on burrows and 0.87 kg d-1 (78 %) were produced on the steppe. Since burrows occupied 0.095 ha within this hypothetical hectare of mountain steppe, burrow biomass production per area was 2.53 kg ha-1d-1. Burrow biomass production per area was thus greater than steppe biomass production per area (0.96 kg ha-1d-1), although this difference was not significant in the analysis of variance for August 2002 (Table 5.4 on page 114). Pikas harvested 0.12 kg d-1 on this hectare mountain steppe, 0.022 kg d-1 (20 %) on burrows and 0.094 kg d-1 (80 %) on the steppe. Pikas were thus using the burrow and steppe habitat similar to the biomass production in the habitats. Biomass on burrows was reduced by 9 %, and biomass in the steppe habitat was reduced by 11 %, which means that pika removed relatively more from steppe plots. This difference was not significant in the analysis of variance, since there was no significant interaction between pika grazing and habitat (Table 5.4). In total pikas removed 11 % of the biomass produced on this hypothetical hectare mountain steppe. Until July 2003 3.45 kg d-1 dry plant biomass were produced (Table 5.7). Bur- rows contributed 0.47 kg d-1 (14 %) and steppe 2.98 kg d-1 (86 %). Burrow production per area (4.96 kg ha-1d-1) was higher than steppe production per area (3.30 kg ha-1d-1). Pikas removed 1.30 kg d-1, 0.17 kg d-1 (13 %) from burrow and 1.12 kg d-1 (87 %) from steppe habitat. They were thus using burrow and steppe habitat proportionally to the biomass produced. Pikas reduced biomass on burrow plots by 35 %, while steppe biomass was reduced by 37 %. In total pikas removed 37 % of the biomass produced up to July 2003. Above-ground plant biomass production was highest in August 2003 with 8.06 kg d-1 (Table 5.7), from which 1.38 kg d-1 (17 %) were produced on burrows and 6.71 kg d-1 (83 %) were produced in steppe. Again, production per area (14.19 kg ha-1d-1) was higher on burrow plots than on steppe plots (7.41 kg ha-1d-1). Pikas removed 0.71 kg d-1 on this hypothetical mountain steppe, 0.11 kg d-1 (15 %) on burrows and 0.60 kg d-1 (85 %) in steppe. Again the pattern of using burrow and steppe habitats was proportional to the production on these habitats. Burrow biomass was reduced by 8 % and biomass in steppe was reduced by 9 %. In total pikas removed 9 % of the biomass produced.

118 5. ECOSYSTEM IMPACT 5.1. PRODUCTION AND BIOMASS REMOVAL

Burrow July Burrow August 25 25

no pika grazing 20 20 pika grazing 15 15 10 10 50

] no pika grazing 2 5 pika grazing 5 m 40 g 0 0

30 All Agr Rem Sti All Agr Rem Sti 20 Steppe July Steppe August 25 25 10 Oven dried biomass [ 20 20 0 15 15 August 2002 July 2003 August 2003

10 burrow steppe burrow steppe 10 burrow steppe 5 5 0 0 All Agr Rem Sti All Agr Rem Sti

Figure 5.3: Standing crop estimated for an initial cover of 8 % for the different harvest times (July and August 2003), habitats (burrow and steppe), plant groups (Agr: Agropyron cristatum, All: Allium spec., Rem: group of remaining plant species, Sti: Stipa spec.), and contrasting pika-grazed to non-pika grazed plots. For significant effects see Table 5.6 on the next page.

Relative to the area covered by burrow habitat, burrows always produced more above-ground plant biomass than steppe habitat. This effect however was significant only for the contrast between July and August 2003 (Table 5.4). The ratio of burrow to steppe productivity increased from 1.5 in July to 1.9 in August 2003, while precipitation increased from 31.8 mm to 62.2 mm. Thus, an increase of 96 % in precipitation lead to an increase of 27 % in the ratio of burrow to steppe productivity. At the same time, burrow and steppe habitat were harvested with the same intensity. Although the proportional removal of plant biomass by pikas (Table 5.7 on page 121) was consistently lower on burrows than on steppe plots, this difference was not significant (Table 5.4 on page 114).

Discussion The inﬂuence of site conditions on plant biomass productivity is predicted to be limited in arid non-equilibrium systems (Fernandez-Gimenez and Allen-Diaz, 1999; Wesche and Retzer, 2005). In another study we show that burrows are enriched in nutrients at the research site (Wesche et al., submitted). Mongolian pikas relocate nutrients from the steppe matrix onto their burrows by carrying livestock dung to the burrows and by concentrating their own excretions and plant biomass in form of harvested plants in their burrows (Retzer, 2004). In the present study, burrow productivity was higher than steppe productivity when more water was available (Figure 5.2,

119 5.1. PRODUCTION AND BIOMASS REMOVAL 5. ECOSYSTEM IMPACT

Table 5.6: The significant effects detected by the ANOVA of above ground plant biomass explained by pika grazing (pika), plant species group (group: All. spec.=Allium spec., Agr. cri.=Agropyron cristatum, Stipa spec., rem= remaining plants), burrow or steppe habitat (hab), and harvest time (time). Effects were corrected for vegetation cover, compare Figure 5.3 on the preceding page and Table 5.5 on page 116.

Effect description pika Pooling all other factors, above-ground plant biomass was lower on plots under pika grazing. group Pooling all other factors, above-ground plant biomass of Al- lium spec. was highest, followed by Agropyron cristatum, the group of remaining plants, and then Stipa. All. spec. pika Under pika grazing, Allium spec. showed consistently more × above-ground biomass than in plots not grazed by pikas. Agr. cri. pika above-ground plant biomass of Agropyron cristatum was re- × duced more than the other plant species under pika grazing. hab Agr. cri. pika The grazing effect for Agropyron cristatum was more evident × × on burrows than on steppe plots. This means at the same time that in the absence of pika grazing Agropyron cristatum showed even higher biomass on burrows than in the steppe. hab rem pika Above-ground plant biomass of the remaining plants was × × higher under pika grazing on burrows, while above-ground biomass was reduced on steppe plots. time Pooling all other factors, above-ground biomass was higher in August than in July. hab time Biomass was even higher on burrow plots in August. × pika time In August the grazing effect was less evident. × rem time When contrasting Stipa spec. and the group of remaining × plants, Stipa spec. biomass was less in August than in July. pika time hab Pooling all species groups, the grazing effect on burrows in × × August was negligible, while it was evident on steppe plots. pika time rem Above-ground plant biomass of the group of remaining plants × × increased in August under pika grazing, while it was reduced in July. All. spec. pika time hab In August, the increase of Allium spec. biomass under pika × × × grazing was higher on steppe plots than on burrow plots which means that Allium spec. was especially avoided by pikas on steppe plots. p. 115). This shows that site conditions matter and that system behaviour varies in time.

Agropyron cristatum Most of the increase in biomass on pika burrows under higher moisture conditions was due to the grass Agropyron cristatum (Figure 5.3, p. 119). This means that although productivity of Allium spec. was highest when

120 5. ECOSYSTEM IMPACT 5.1. PRODUCTION AND BIOMASS REMOVAL

Table 5.7: Above-ground plant biomass production and removal by pikas on a hypothetical hectare of mountain steppe. Burrows occupy 0.095 ha of the surface area, while steppe habitat occupies 0.905 ha. Approximate time intervals were 30 days up to August and 75 days up to July, assuming that growth started in the middle of May.

Sample date Habitat Production Biomass removal by pikas [kg/d] [kg/d] [% of production]

August 2002 Burrow 0.24 0.02 9 Steppe 0.87 0.09 11 July 2003 Burrow 0.47 0.17 35 Steppe 2.98 1.12 37 August 2003 Burrow 1.35 0.11 8 Steppe 6.71 0.60 9

pooling all data, A. cristatum showed highest productivity when pikas were excluded and even more so on burrow plots. So, apart from being an important fodder plant for livestock (Retzer, 2004), A. cristatum was preferentially harvested by pikas in the present study. Other pika species select plants as a function of the distance to their talus in North America (Ochotona princeps, Huntly, 1987; Mcintire and Hik, 2002). In Asia, Smith and Foggin (1999) state that burrows are grazed more intensively by pikas (O. curzoniae). This behaviour is always set into the context of risk-aware foraging, as pikas are less prone to predation when they can hide in their talus or burrows (Lutton, 1975). However, in the present study there was no evidence that burrows were preferred grazing grounds when pooling species groups. The habitats were rather used according to the amount of biomass produced on them. Discriminating between plant species groups showed that A. cristatum was indeed preferentially removed from burrow plots by pikas, while the group of remaining plants was avoided. Moreover, when overall plant biomass was larger, pikas removed relatively more A. cristatum than other plants. This grass effectively was the only plant grazed in August 2003 (Figure 5.3, p. 119), when water availability and plant productivity was highest. The less water and standing crop was available, the more uniform was the influence of grazing. Retzer (2005) shows that under drought conditions even Allium spec. is removed by pikas, while this group of species always showed increased productivity under influence of pika grazing in the present study. Moreover, A. cristatum is the only plant species showing significantly higher nutrient content on burrow habitat (Wesche et al., submitted). Thus, the preference for this species exhibited by pikas is less related to predation risk, but to the quality of the plants. Mongolian pikas do not behave risk-aware in plant selection, but taste-aware. Finding that pikas avoid the group of remaining plants on burrow plots (Figure 5.3, p. 119) is in accordance with taste-aware foraging: the group of remaining plants includes more ruderal plants on the burrows, which are less palatable (Wesche et al., submitted).

Selective grazing and plant species composition Herbivore inﬂuence on plant species composition decreases along a hygric gradient (Olﬀ and Ritchie, 1998). Accord-

121 5.1. PRODUCTION AND BIOMASS REMOVAL 5. ECOSYSTEM IMPACT

ingly, in the present study, plant species composition does not reflect the selectivity of pika grazing. Although Allium spec. was consistently avoided by pikas throughout the experiment, while Agropyron cristatum was consistently removed most compared to all other plant groups, A. cristatum dominates burrow and steppe habitat. Additionally, although it is grazed even more intensively on burrows, it reached higher cover values on burrows than in the steppe. This shows that even though Allium spec. and the group of remaining plants can increase production due to pika grazing, other mechanisms must be responsible for controlling plant species composition. Important factors include disturbance by burrowing or soil nutrient contents (Wesche et al., submitted). When studying influences on the composition of vegetation, factors affecting below-ground competition should also be regarded (Lavrenko et al., 1993; Titlyanova et al., 1999). The findings of the present study are in accordance with studies showing that competition intensity between plants decreases with decreasing productivity (Sammul et al., 2006). Higher cover values and quality of a preferred forage plant (Agropyron cristatum) can be seen as an example of positive feedback engineering by pikas (Jones et al., 1997). Livestock and other wild herbivores also profit from higher A. cristatum cover. Larger herbivores have been observed to graze preferentially on pika burrows, especially in spring (own observation and Retzer, 2004). Furthermore, pika grazing may influence plant species composition in other ways not studied here. Pikas in general are known for their urge to collect (Millar, 1972) and many species collect especially flowers (Smith et al., 1990). The individuals on the study site have been observed to prefer eating the flowers of legumes (Astragalus spec. and Oxytropis spec.), and to collect all sorts of colourful things on their burrows. This may have an influence on the reproductive success of plant species reproducing sexually. Since sexual reproduction is rare in the research area (Wesche et al., in press), another behaviour may be even more influential: Whitford and Kay (1999) state the importance of foraging pits as nutrient rich germination sites. The observed pikas made foraging pits up to about 12 cm in depth (unpublished data). Some of these foraging pits were subsequently used by ants and filled with fine grain soil. Ants themselves qualify as ecosystem engineers, creating hot-spots in the ground and maintaining ecosystem heterogeneity (Jouquet et al., 2005). The impact of foraging pits of pikas on the ecosystem has not been studied yet.

Comparison with other studies on the research area Wesche and Retzer (2005) find a difference between grazed and ungrazed plots of 26 % in August 2002 and of 25 % in August 2003 on 4 larger enclosures on steppe habitat in the same study area. This corresponds well to the percentages found both on burrow and steppe plots in this experiment, since their data summarise productivity and removal from the beginning of the growth period. Assuming a removal rate of about 35 % for three months and of 10 % for one month would result in a removal rate of 29 % for the data from this experiment. But their estimate for relative biomass removal in the drought year 2001 (78 %) is far higher. In 2001 pika numbers indeed were higher than in 2002 (Chapter 3, p. 47), while biomass production was low due to the year of drought. Grazing pressure must have been exceptionally high in this year. Even an 20 % biomass-removal rate is high compared to that effected by other small mammals in the Eurasian steppe region,

122 5. ECOSYSTEM IMPACT 5.1. PRODUCTION AND BIOMASS REMOVAL

Lavrenko et al. (1993) giving an average of 6 % biomass removed by small herbivores. They say that biomass removal rates of over 70 % are detrimental for the ecosystem. The high biomass removal rates in 2001 thus stress the severeness of the drought.

Harvesting Non-equilibrium theory predicts that not herbivory but climatic conditions control the productivity of arid systems (Fernandez-Gimenez and Allen-Diaz, 1999; Wiens, 1984). In the present study, the removal of plant biomass by pikas could not keep up with the varying amount of plant biomass produced under different moisture conditions: biomass production reflected the amount of precipitation even under pika grazing (Figure 5.2, p. 115). At the same time, the amount of biomass removed by the pikas was several times higher in the year 2003 than in 2002 (Table 5.7, p. 121), although pika numbers were about the same (Table 5.1, p. 110). Moreover, the relation of biomass removed to the biomass available was surprisingly constant for the two Au- gust samples from 2002 and 2003 (Table 5.7, p. 121). Most of the biomass harvested in the year 2003 was probably not eaten, but stored in the burrows. It is questionable, if the pikas can consume all these plants during winter, since it must have been much more than in the year before. Ochotona daurica stores up to 56 kg ha-1 plant biomass annually in eastern Hangay (Lavrenko et al., 1993). Storing food can thus act similar to key resources (Illius and Connor, 1999) for the pikas, mitigating the effects of variable biomass production in a non-equilibrium system. The plant biomass stored in the burrows and not used up during the winter can then be used as supplement fodder during adverse conditions in the following growing season. If the following growing season has favourable conditions for plant growth, the stored plant material decays within the burrow, further enriching the burrow with nutrients. The intensity of harvesting changes during the year. The observed pikas removed more biomass proportionally in July 2003 than in August 2002 and 2003. This was not limited by the amount of biomass, since in the year 2003 there was enough plant biomass available (Table 5.7, p. 121). In another study we show that harvesting is intense during September (Retzer and Nadrowski, 2002). Other studies also show that there are peak harvest events in autumn (O. princeps, Mcintire and Hik, 2002). Pika caching behaviour continues even in winter (own observations, and Kawamichi, 1968; Retzer, 2004). Finding reduced harvest intensity during late summer has to be seen in the context with the territorial behaviour of the pikas. They probably shift priorities to territory protection during late summer.

Livestock grazing and pika–livestock interaction Although this experiment could not detect livestock grazing (main eﬀects and interactions were only marginally signiﬁcant, Table 5.4, p. 114), livestock grazing is observed on the larger enclosures on steppe environments described by Wesche and Retzer (2005). Livestock grazing on those four plots is comparable to pika grazing and plots grazed by both groups are reduced to even lower standing crop. Not detecting livestock grazing by the present experiment probably was caused by the size of the enclosures. While the larger enclosures measure 4 4 m2, large enough for livestock to step into the enclosure, the small × fences used in this experiment measured 0.26 m2. Livestock could not step easily into the small fences and might just have not bothered about them and moved on, especially since there was enough plant biomass available at least in the year 2003. The

123 5.2. PIKA DENSITIES ON REGIONAL SCALE 5. ECOSYSTEM IMPACT

percentages grazed in the years 2002 and 2003 by livestock and pikas respectively sum up to the percentage grazed on unfenced plots in Wesche and Retzer (2005), suggesting that in these years there might have been less to no competition between the grazing groups, although there is intense competition under drought conditions (Retzer, 2004). The intensity of competition thus varies with water and biomass availability.

Conclusion Both, removal of plant biomass and burrowing activity, effect ecosystem qualities in the mountain steppe. While plant biomass removal leads to differences in the productivity of plant species groups, it does not effect the composition of the vegetation. Burrows show higher productivity of an important fodder plant (Agropyron cristatum), and can thus be seen as beneficial also for livestock and other herbivores. At the same time, increased burrow productivity cannot outweigh the effect of biomass removal by pikas when balancing production and removal in the system (Table 5.7, p. 121). Nevertheless, pika burrows are places were nutrients are concentrated and conserved. This is even more important, since human usage depletes the system of its nutrients by using dung of livestock as fuel source (Stumpp et al., 2005). Competition between livestock and pikas might be less severe or even absent under favourable conditions for plant growth, but the removal of biomass by pikas is nevertheless high in comparison with other small mammal species. Pika harvesting behaviour leads to large amounts of biomass removed, even if pika densities are low. However, pikas cannot keep up with the variability of plant productivity. The system behaves in accordance with non-equilibrium theory, since climatic conditions and not pika grazing control the amount of standing crop present.

5.2 Pika densities along an altitudinal transect

Chapter 4 on page 57 shows that pika density is mainly controlled by the availability of burrows. But which factors control the density of burrows over a longer period of time? Climatic conditions change along altitudinal gradients and Eurasian steppes show marked diﬀerences in species composition along altitudinal gradients (Lavrenko et al., 1993). Since precipitation usually increases with altitude, moister steppes can be found in higher altitudes. At the same time, ecosystem behaviour changes along moisture gradients, from precipitation controlled non-equilibrium systems to herbivore controlled equilibrium systems (Ellis and Swift, 1988; Fernandez-Gimenez and Allen-Diaz, 1999; Wiens, 1984). Additionally, small mammal densities can increase when the range is overgrazed by livestock (Samjaa et al., 2000; Zhong et al., 1985). The change from socialist to market economy in the 1990s led to increasing livestock numbers and a growing concern about the limit of a carrying capacity of the steppe environment in Mongolia (Muller¨ , 1999). Livestock numbers also increased in the South Gobi Aimag, where the research area is situated (Retzer, 2004). This section asks, whether altitude or livestock densities have an eﬀect on pika density. It then uses the results on biomass removal by pikas and burrow productivity to estimate the impact of pikas on the ecosystem for varying altitudes and livestock densities.

124 5. ECOSYSTEM IMPACT 5.2. PIKA DENSITIES ON REGIONAL SCALE

Table 5.8: Ratio of burrow to steppe productivity estimated for standing crop on steppe habitat and precipitation on diﬀerent altitudes. Data on precipitation and standing crop are from Retzer et al. (in press) and Retzer (2004). Precipitation data come from October 2000 to October 2001, and maximum estimated standing crop within the summer of 2001. Numbers in parenthesis are own interpolation.

Altitude [ m asl] 2000 2200 2300 2400 2500 2600 2800

Precipitation [mm] 42 72 (95) 117 (125) 132 97 Standing crop [ kg ha-1] 43 43 193 260 (230) 200 320 Ratio burrow/steppe 0.66 1.40 1.5 1.59 1.64 1.67 1.52

Methods Pika and livestock densities were investigated using line-transects at different altitudes in the Duund Saikhan and Zuun Saikhan mountain ranges (Section 2.1, p. 13). Altitudes ranged from 1900 to 2800 m asl. Number of burrows were used as a proxy for pika density and dung deposits as a proxy for livestock density. Burrows of pikas and dung deposits of large herbivores (sheep, goats, ibex, argali, cattle, horses, camels) were counted within a distance of 10 m on each side of the transect, resulting in a corridor of 20 m width. Thirty-one transects were used in the analysis, with lengths ranging from 50 to 1490 m. A total length of 21 362 m was investigated, of which 3 934 m were located in the Zuun Saikhan and 17 428 m in the Duund Saikhan. The data was analysed using a generalised linear model with Poisson-errors. The number of burrows (b) was the response variable, while altitude (a), dung/ha (d) were used as explaining variables. The area (ar) of the part-transect strip was given as an additive covariate. Altitude was analysed with linear, quadratic, and cubic terms. The formula of the general model was b ar + d (a + a2 + a3). (5.2) ∼ · The general model was simplified using the step-function of R, where terms are dropped one at a time when AIC values of the reduced model are smaller than those of the original model (R Development Core Team, 2004). Models differing in less than 2 in AIC were additionally compared with a likelihood ratio test. If the models did not differ significantly the reduced model was chosen. The findings of plant productivity, pika forage, and pika densities were summarised to estimate productivity and forage along an altitudinal transect at low, median, and high dung densities. Data on precipitation and steppe productivity originated from an altitudinal transect at the study site (Section 2.1, p. 13). Both generally increase with altitude at the research station (Retzer, 2004; Retzer et al., in press). The model was based on the following assumptions:

1. Biomass removal is proportional to the production of biomass. Table 5.7, p. 121 shows that with a comparable number of pikas and varying amounts of standing crop, pika biomass removal was about 35 % of standing crop up to July and 10 % from July to August. This assumption ignores the year of drought, where biomass removal rates were higher (Wesche and Retzer, 2005). 2. The rate of biomass removal is proportional to the number of burrows. The number of burrows was taken as proxy for the number of pikas in this case. Analysis

125 5.2. PIKA DENSITIES ON REGIONAL SCALE 5. ECOSYSTEM IMPACT

of survival rates and pika densities showed that the observed population of Mon- golian pikas was density dependent. Density is probably limited by the number of territories available, which in turn is dependent on the number of burrows (Section 4.5, p. 101). The rate of biomass removal measured at the research camp (Table 5.7, p. 121, see the ﬁrst assumption) was calibrated by the number of burrows measured for the height of the research camp using the altitudinal transect (this section).

3. Burrow and steppe exhibit diﬀerent productivities. The relation of burrow to steppe productivity varied with precipitation. An increase of 96 % in precipitation lead to an increase of 27 % in the ratio of burrow to steppe productivity in summer 2003 (Section 5.1, p. 109). Since 2003 was a relatively moist year, while the precipitation data for the altitudinal transect originated from a year of drought, the lower of the two ratios was used for the altitude of 2300 m asl. Ratios for the other altitudes were then extrapolated proportional to the increase or decrease of precipitation for the altitudinal transect in relation to the precipitation at 2300 m asl. (Table 5.8, p. 125).

4. The area covered by a single burrow remains constant over an altitudinal gradient, although burrow numbers change with altitude and livestock density as estimated by the model above. The number of burrows at a given altitude than determines the proportion of the area covered by burrows in the steppe environment.

Formulae Since burrows occupy between 7 and 12 % of the surface in our study region, the mean of 9.5 % was assumed to be the area covered by burrows at the study site (Wesche and Retzer, 2005). Estimated burrow density for median livestock dung densities at 2300 m asl was 24.7 burrows per hectare using the linear model above. This results in an estimated area of 38.46 m2 per burrow, which is a fair estimate given that the median burrow size on the capture recapture site was 38.1 m2 (Section 2.4, p. 26). 2 With 38.46 m as area for a single burrow (Ba), the area occupied by burrow (Ab) or steppe habitat (As) at a given altitude is a function of burrow density (nb). On one hectare, the area occupied by burrows is Ab = nbBa, while the area occupied by steppe habitat is A = 1 n B . s − b a Steppe above-ground biomass production per area (ps) relies on data from the altitudinal transect at the study site in 2001 (Retzer et al., in press, Table 5.8). The ratio of burrow to steppe productivity per area ( pb ) is assumed to change with precip- ps itation (Table 5.8, Assumption 3 above). Higher altitudes receive more precipitation in the study region (Retzer et al., in press; Wesche et al., in press). As an estimate for the diﬀerences across altitudes, the sum of precipitation from October 2000 to October 2001 from the same altitudinal transect was used (Retzer et al., in press). Above-ground biomass production (Ps) in steppe habitat was thus given as Ps = psAs, while above-ground biomass production on burrows (P ) was P = p pb A . b b s ps b Total above-ground biomass production on one hectare (Pt) was thus calculated as the sum of biomass produced on burrow and steppe habitat from the middle of May to August, given the number of burrows at a given altitude (nb), the area of a single

126 5. ECOSYSTEM IMPACT 5.2. PIKA DENSITIES ON REGIONAL SCALE 2800 2800 2600 2600 2400 2400 altitude [m asl] altitude [m asl] 2200 2200 Altitude [m asl]

low dung density median dung dens.

2000 high dung density 2000 standard errors

0 10 20 30 40 0 10 20 30 40

1 burrow density [burrows/ha] burrow density [burrows/ha] Burrow density [burrows/ha]

Figure 5.4: Burrow densities along an altitudinal gradient from 1900 to 2000 m asl. Lines show the estimated number of burrow at diﬀerent dung densities, while points show the measured burrow densities [burrow/ha]. The left plot shows burrow densities at low and high dung densities (21.06 and 61.07 dung deposits/ha), while the right plot shows burrow densities for median dung densities (32.45 dung deposits/ha) with standard errors. burrow (B ), and the ratio of burrow to steppe productivity per area ( pb ). a ps

pb Pt = ps(1 nbBa) + ps nbBa (5.3) − ps

Plant biomass removed by pikas was supposed to be proportional to biomass production (Assumption 1 above) and the number of burrows at a given altitude (As- sumption 2 above), with a biomass removal rate of 35 % from the middle of May to August (75 d), and a biomass removal rate of 10 % in August (30 d) at the altitude of 2300 m asl. A rate of biomass removal per burrow (fb) was then calculated using the number of burrows estimated for median dung density by the generalised linear model described above. At a given altitude, biomass removal rate per area (f) can now be calculated as a function of burrow density (nb): f = nbfb. Steppe productivity per area (ps) gives the amount of biomass that would have been produced without burrows. Total above-ground biomass production, biomass removal by pikas, and steppe productivity without burrows were calculated for the number of burrows predicted at low, median, and high dung densities along the altitudinal transect.

Results The general model relating burrow densities to altitude and livestock densities could not be reduced. It had an AIC of 1038.85, with the nearest reduced model differing by 2.47 in AIC. Both models could be compared using analysis of variance, since they were nested. The comparison resulted in significant differences between the models (χ2-test, p = 0.03). Thus, all terms of the model were significant, including the interactions. Figure 5.4 shows the predictions for burrow densities [burrow/ha] for different dung densities. Burrow densities increased with altitude and then decreased

127 5.2. PIKA DENSITIES ON REGIONAL SCALE 5. ECOSYSTEM IMPACT

Table 5.9: Increase of plant biomass production in the mountain steppe when including burrow habitats (increase due to burrows) and net change after biomass removal by pikas along an altitudinal transect and at diﬀerent dung densities (low, median and high). Dung densities were 21.06, 32.45, and 60.18 dung deposits/ha at low, median, and high densities respectively.

Altitude Steppe Increase due to burrows Net change after removal [ m asl] [ kg ha-1] low median high low median high

2000 43 1.00 1.00 1.00 0.96 0.97 0.99 2200 43 1.03 1.03 1.02 0.80 0.81 0.84 2300 193 1.05 1.05 1.05 0.74 0.74 0.73 2400 260 1.05 1.06 1.06 0.76 0.75 0.71 2500 (230) 1.04 1.05 1.05 0.82 0.81 0.77 2600 200 1.03 1.03 1.04 0.88 0.87 0.85 2800 320 1.01 1.01 1.01 0.94 0.94 0.94

again, the maximum was reached between 2330 and 2370 m asl for the given dung densities. Dung densities did not have a strong effect in modulating burrow densities, although the effect was significant. Nonetheless, there was a positive correlation between dung densities and burrow numbers at altitudes above 2300 m asl and a negative correlation at lower altitudes. Table 5.9 lists the changes estimated for above-ground biomass production with burrow and biomass removal by pikas relative to steppe production at low, median, and high dung densities (21.06, 32.45, 60.18 dung deposits/ha). The ratio of standing crop with burrow habitat to standing crop in steppe habitat ranged from 1 to 1.06. At 2300 m asl the ratio was the same for all dung densities, while at altitudes lower than 2300 m asl the ratio decreased with increasing dung densities and for altitudes higher than 2300 m asl it increased with dung densities. Net change in above-ground biomass including plant production on burrows and biomass removal always lead to a decrease of standing crop. The ratio of standing crop including burrow habitat and biomass removal to standing crop in steppe were about constant at 2300 m asl, increased with increasing dung density below that altitude and decreased above that altitude. Differences predicted by dung densities were smaller than differences predicted by altitudes.

Discussion Increasing burrow density with altitude implies that the long-term availability of forage inﬂuences the size of pika territories on the time scale of burrow lifetimes. In the present study, burrow density reached a maximum at about 2370 m, which corresponds with the pediment angle, separating the pediments from the mountains. Highest burrow densities are thus found on the highest pediments, just before the landscape changes to the ragged mountain relief. This pattern is in accordance with the distribution of mountain steppes. While they cover most of the area on the upper pediments, they are less prominent in the mountains themselves. Within the mountains, there are three main plant communities, those dominated by the dwarf shrub Artemisia santolinifolia, scrub composed of Juniperus sabina, and mountain steppes (Wesche et al., 2005). Although pikas can be found in all three plant communities,

128 5. ECOSYSTEM IMPACT 5.2. PIKA DENSITIES ON REGIONAL SCALE

the results of the present study imply a causal connection between the prevalence of mountain steppe plant communities and burrow density, which may simply be the productivity of the vegetation. The plant communities dominated by Artemisia santolinifolia and Juniperus sabina are characterised by higher soil movement than the mountain steppes (Wesche et al., in press, 2005), so that conditions for plant productivity are better in the mountain steppes. Thus, a lower burrow density within the mountains than at the pediment angle supports the link between the availability of forage and burrow density. Since burrow lifetime probably spans several decades to centuries (Section 2.4.1, p. 26), burrow density may be seen as a measure of ecosystem productivity, which integrates over the available plant biomass in centuries. This measure includes a time lag of one year in the response to conditions not favourable for plant growth due to the harvesting behaviour of the animals (Section 5.1, p. 109). At the same time, the Mongolian pika did not show a response to livestock densities as described for the Brandt’s vole for northern Mongolia and China (Samjaa et al., 2000; Zhong et al., 1985), where voles invade overgrazed pastures. Although there was a significant positive effect of livestock dung densities on pika burrow densities above 2300 m asl, this effect was minor in comparison to changes induced by altitude (Table 5.9, p. 128). On the other hand, there are several methodological diffi- culties connected with the analysis presented here. Livestock densities do not show much differences in the study area (Stumpp et al., 2005; Wesche et al., in press), the whole area is characterised by high grazing intensity of livestock (Retzer et al., in press; Stumpp et al., 2005). Thus, influences of livestock density on pika burrow density may have been missed because there are no areas with low livestock density. Additionally, burrows and dung deposits reflect different time scales. While burrows last for decades to centuries, dung deposits can only indicate livestock densities of the last months. Nonetheless, the findings of the life history strategy adopted by pikas (Section 4.5, p. 101) are in accordance with the conclusion that pika densities are virtually independent of livestock densities. The observed pikas showed characteristics of K-selected species, implying that they have relatively constant population densities controlled by the availability of territories. Although pika densities varied in the three summers observed, they did not vary much in comparison to other pika species which do not control densities by defending territories (Table 4.19, p. 102). Finding only small effects of livestock densities on pika burrow densities is in accordance with density dependant populations dynamics controlled by the availability of territories. There was a significant, though very small, positive effect of livestock densities on pika burrow densities above 2300 m asl, and a negative effect below. Similar to the increase and decrease of burrow density, this change in system behaviour corresponds with the pediment angle. Negative correlations of pika densities and livestock densities have been shown by Komonen et al. (2003) for Ochotona daurica in eastern Mongolia. Finding a positive correlation in the mountain indicates that conditions are different in the mountains than on the pediments. Since both water and forage availability are higher in the mountains, the system behaviour may tend more to those of equilibrium systems (Fernandez-Gimenez and Allen-Diaz, 1999). Under equilibrium conditions however, grazing exerts more influence on the vegetation than under non- equilibrium conditions. This is in accordance with not finding indicators of overgrazing in the dryer plant communities of the GGS, but in the moister ones (Miehe, 1996, 1998; Retzer, 2004). Although including pika burrows in the balance increased the productivity of the

129 5.2. PIKA DENSITIES ON REGIONAL SCALE 5. ECOSYSTEM IMPACT

vegetation, biomass removal by pikas always surpassed this increase. Thus, although a conservation of nutrients is important in a system were nutrients are systematically removed by humans (Stumpp et al., 2005; Wesche et al., submitted), simply including higher burrow productivity doesn’t change the conclusions drawn from steppe plots alone. Mongolian pikas and livestock must be regarded as competitors at least under drought conditions (Retzer, 2004).

130 5. ECOSYSTEM IMPACT 5.3. SUMMARY

5.3 Summary

Burrows showed higher productivity than the steppe when water availability was higher. An increase of precipitation by 96 % from 31.8 mm to 62.2 mm lead to an increase of burrow productivity of 186 % from 4.96 to 14.19 kg ha-1d-1, while steppe productivity increased by only 125 % from 3.3 to 7.41 kg ha-1d-1. The grass Agropyron cristatum profited most from the burrow habitat, as indicated by higher productivity when pikas were excluded and by the dominance of this species on burrow habitats. Since A. cristatum is an important fodder plant, pika burrowing activity can be considered beneficial for livestock grazing. The Mongolian pika preferentially removed biomass of A. cristatum from burrow and steppe habitat. Biomass removal by pikas was high, ranging from 0.11 to 1.29 kg per hectare and day. Although the amount of plant biomass removed per pika was different, the proportion removed of standing crop was similar in August 2002 and 2003. Pikas removed about 10 % of standing crop during August and 35 % of standing crop during May to July (Table 5.7, p. 121). Large amounts of the harvest were probably stored in the burrows and not immediately consumed. The system thus behaved according to non-equilibrium theory, since water availability and not pika grazing controlled the amount of standing crop. Although livestock grazing could not be detected by the present analysis, this was probably due to the size of the exclosures and not due to the absence of livestock grazing. Difference in altitude had a strong effect on pika burrow density. This is probably due to a gradient in productivity. Maximum burrow densities were reached at the pediment angle separating the gently sloping pediments from the ragged mountains. However small, there was an effect of livestock densities on burrow density, indicating a change in system behaviour at the pediment angle. Livestock and pika density were positively correlated in the mountains, while they were negatively correlated on the pediments. Increased burrow productivity could not be expected to outweigh the effect of biomass removal by pikas, even if increasing precipitation with altitude is included.

131 5.3. SUMMARY 5. ECOSYSTEM IMPACT

132 Chapter 6

General Discussion

When assessing whether the Mongolian pika (Ochotona pallasi pricei) might appear as “pest” species, the life history strategy of the species can set the frame for possible scenarios of the species’ impact on the environment (Putman, 1989). In the case of the present study, the Mongolian pika shows life history traits which reduce its chance to appear as small mammal pest (Section 6.1). Additionally, when assessing possible repercussions of control, especially the pikas’ impact on standing crop and the composition of the vegetation through burrow maintenance rather imply that they are an indispensable part of ecosystem functioning in the present mountain steppe (Section 6.2, p. 137).

6.1 Life history strategy and habitat quality

The present study shows that in relation to other pika species the Mongolian pika can be considered a K-strategist (Table 4.19, p. 102). Survival of the observed individuals was density-dependent, while climatic conditions, including winter and a drought, did not have an eﬀect on survival. Additionally, number and size of litters as well as ﬂuctuations in population density were low compared to other pika species (Chapter 4, p. 57, Smith, 1988). The group of species in pikas exhibiting life history traits similar to an r-strategist is named “burrowing” pikas, or “steppe-dwelling” pikas in Smith et al. (1990) and Smith (1988). The underlying reason for the evolution of the two life history patterns is seen in the predictability of the environment, similar to the r/K-concept (Section 1.3, p. 3). In this concept, the investment into present survival is expected to be low in an unpredictable environment, since survival is governed by the environment and thus also unpredictable. The link between burrowing behaviour and exhibiting r-type life history traits is rather correlative than causative: open ecosystems are often subject to a higher variation in moisture availability, since they cannot sustain forests (Schultz, 1995). At the same time, burrows present means to hide from predators or from adverse climatic conditions for organisms living in open environments (Kinlaw, 1999). Finding a burrow-dwelling pika in a variable environment to exhibit life history traits conforming with the K-type of life history strategies contradicts the link between habitat quality and life history strategy as suggested by (Smith, 1988). It might imply that (a) the species has not evolved in the steppe environment where it occurs today, or (b) that the studied steppe environment is not as unpredictable as might be expected,

133 6.1. LIFE HISTORY STRATEGY GENERAL DISCUSSION

or (c) the r/K-concept is not suﬃcient to describe the inﬂuence of the environment on the evolution of life history traits.

The Mongolian pika is not adapted to open steppe environments. Although it uses burrows, it does not invest much time and eﬀort into the construction of burrows. It does not have the long claws (Zevegmid, 1975) characteristic for the other burrowing pika species (Smith, 1988), indicating that it might not be adapted to burrowing. During the present study, the Mongolian pikas did not dig much. In the three years of observation, there was no change in the location and size of the burrows, burrows were estimated to last at least 120 years (Section 2.4.1, p. 26). Thus, Mongolian pikas use the same burrows for generations, which frees them from investing time into burrow construction. Additionally, they used available structures to hide within. In the mountainous landscape Mongolian pikas used rock crevices and Juniperus-scrub, while individuals of the same species constructed burrows on grass dominated slopes (own observation). Another argument supporting that Mongolian pikas did not evolve in a steppe environment is a karyological study of eight taxa of pikas, which suggests that the steppe dwelling pikas (Ochotona pusilla, O. daurica, and O. pallasi pricei) might have entered the steppe environment at three independent times (Vorontsov and Ivanitskaya, 1973 quoted in Yu et al., 1997). Yu et al. (1997) suggest that the genus experienced a rapid radiation in the late Pliocene and early Pleistocene, caused by the uplifting of the Ti- betan (Qinghai-Xizang) Plateau. This led to colder and, especially in the rain-shadow of the Himalaya massif, dryer conditions. Climatic oscillations with glacial and inter- glacial transitions, but without a uniﬁed ice-sheet, resulted in continual habitat shifts. Pikas probably originated from the cold, arid steppes created during colder phases with glaciation. The research area itself has been subject to glaciation (Wesche et al., in press). Thus, Ochotona pallasi pricei might have entered the steppe environment later than the other burrowing species. A further argument for this is that the species is found in poorer habitats if it occurs in sympatry with other species, as shown for O. pusilla (Shubin) and O. daurica (Proskurina et al., 1985). Within the research area, O. daurica was found only in places with additional water supply, such as Kobresia mats or saline meadows, while O. p. pricei occurred on the poorer mountain steppe habitats (Section 2.3, p. 25).

The studied steppe environment is not as unpredictable as might be expected. On the one hand the ecosystem processes studied were essentially governed by non- equilibrium dynamics: standing crop was controlled by precipitation and not by herbivore grazing both on burrows and in the steppe, and the composition of the vegetation was not inﬂuenced by pika grazing (Section 5.1, p. 109). Variability of climatic conditions and standing crop on steppe plots in the research area is also nicely demonstrated by Wesche and Retzer (2005). Although individual mortality of pikas in the present study was not aﬀected by drought conditions (Section 4.2.2, p. 81), population density was lower after the year of drought (Chapter 3, p. 47). On the other hand, population densities of the Mongolian pika did not vary as much as observed in other pika species (Table 4.19, p. 102), indicating that the variability

134 GENERAL DISCUSSION 6.1. LIFE HISTORY STRATEGY of the environment is not perceived as such by the Mongolian pika. The mountain ranges are indeed much less affected under drought conditions than the surrounding pediments (Retzer, 2004). The present study showed a shift in ecosystem behaviour at the pediment angle, where pika burrow densities were highest and the correlation between livestock density and burrow density changed from a negative one on the pediments to a positive one in the mountains (Section 5.2, p. 124). Nonetheless, even if the mountain ranges are less governed by non-equilibrium conditions, they still represent a highly variable environment, subject to high interannual fluctuations in precipitation and primary productivity (Retzer, 2005). A mechanism further mitigating the severeness of the environmental variability is the territorial behaviour of the species, together with its caching behaviour. Although several years favourable for plant growth preceeded the year 2000, pika density did not exceed one individual per burrow on the trapping site (Chapter 3, p. 47). By defending territories, individual pikas are able to cache enough food to last not only through the winter, but also through the following summer, since individual survival was not affected by the drought conditions (Section 4.2.2, p. 81). In the present study, individuals in the year 2003 removed more biomass from the plots, than those in 2002, although the proportion of removed biomass in relation to standing crop was the same (Section 5.1, p. 109). Thus, in a favourable year, a large surplus of biomass can be stored in the burrows. Following this argument resurrects the initial link between an r/K dichotomy of life history traits and unpredictable environments: the mountain steppe does not present an unpredictable environment for the Mongolian pika.

The r/K-dichotomy is not sufficient to describe the influence of the environment on the evolution of life history traits. The r/K-concept is only a special case of a more general classification of habitat quality, linking the effect of present reproduction on the residual reproductive value for adults with the effect of number on the size of offspring (Begon et al., 1996). In the present case of the Mongolian pika, the relatively low investment in present reproduction enables higher survival of an individual, which again increases the chance to reproduce later, conceptualised in the residual reproductive value. At the same time, higher survival of the juveniles from the first litter indicates that investment in many young is less rewarding than investment in the first few young (Table 4.19, p. 102). In a review Stearns (1977) finds that of 35 thorough studies, 18 conform to the r/K scheme while 17 do not. Especially modular organisms stand apart. Modular growth gives the genet the potential to increase in size exponentially by increasing the number of modules. Delayed reproduction then does not necessarily delay population growth. This may account for the rather high frequency of species with clonal growth (and potentially infinite life) in r-selecting habitats (Begon et al., 1996). For example, most of the dominant plant species in steppe habitats are perennial (Coupland, 1992a), which can be interpreted as investment in residual reproductive value. This is a life history trait conforming with a K-strategy. At the same time, they produce many seeds, which is a life history trait conforming with an r-strategy. Other concepts try to link groups of species and not life history traits to habitat qualities, as for example the triangular scheme developed for plants by Grime et al. (1988), where plant species are grouped according to their tolerance to stress (S), to disturbance (R for ruderals),

135 6.1. LIFE HISTORY STRATEGY GENERAL DISCUSSION and their competitive ability (C). In the present study, an environment governed by non-equilibrium dynamics hosts a species with K-type life history traits because of behavioural characteristics of this species, namely the aggressive behaviour towards neighbours when defending territories. The aggressive behaviour of Ochotona pallasi is described by Proskurina et al. (1985). For the observed population in the present study, Monkhzul (2005) documents territorial behaviour. Section 2.6 on page 42 showed that median movement distances of individuals correspond with nearest neighbour distances between burrows. The question of finding density-dependence in a non-equilibrium environment then narrows down to the question how the species could preserve its territorial behaviour in the non-equilibrium environment. The other pika species found in the research area occurs in more productive habitats, with higher water availability, and should thus experience less variation in the amount of standing crop. Following the argument above, Ochotona daurica should also exhibit aggressive and territorial behaviour. However Smith et al. (1990) list the species as a burrowing one, with the life history traits leading to highly fluctuating population densities. Ochotona daurica lives in pairs and shows affiliative behaviour more frequently than does Ochotona pallasi. At the same time, it has been made responsible for pasture degradation (Zhong et al., 1985). In the research area, burrows of O. daurica are closer to each other in some places, including Yoliin Am (Figure 2.2, p. 14), a major tourist site in the GGS featuring among other things a permanent water source (own observation). The concern about increasing pika densities probably originated from observations at this site (Section 2.3, p. 25). Additionally, O. daurica is considered a sibling species to the well studied Ochotona curzoniae, which belongs to the group of burrowing species showing less aggressive and territorial than affiliative behaviour. Nonetheless, there may be some support for questioning the classification of O. daurica as more social and less territorial. Tsendzhav (1976) reports densities between 25 and 55 animals (mean 29) at 12 different sites for O. daurica in one study, and densities between 10 and 24 per ha in another (Tsendzhav, 1985). Assuming a mean of 29 individuals on 15.5 burrows results in two animals per burrow. Burrow size is not greater than that of Ochotona pallasi pricei in the present study, with 25.8 m2 and 38.1 m2 respectively (Section 2.4.1, p. 26). This indicates that population densities in O. daurica may be governed by the availability of burrows, similar to Ochotona pallasi pricei. Like O. p. pricei, O. daurica uses available structures as nesting sites rather than digging its own burrows: Schauer (1987) reports that Ochotona daurica in contrast to the other two species studied in the Khentei and Kangai of Mongolia, Ellobius talpinus and Myospalay aspalax, likes to use the burrows of the other species. In burrowing pikas older animals are rare (Smith, 1988, and Table 2.4, p. 24), but studies on O. daurica state high proportions of older individuals (Tsendzhav, 1977, 1985). Additionally, unlike Microtus brandtii but similar to the Mongolian pika, Komonen et al. (2003) show that population densities of O. daurica decrease with livestock density. Thus, a facilitation of O. daurica by overgrazing is questionable. Often O. daurica is cited as a small mammal pest, because it is one of the species occurring on the degraded steppe environments studied by (Zhong et al., 1985). But the species is only found in the least degraded steppes in this study, while it is displaced by Meriones unguiculatus and Microtus brandtii in the steppes with more severe degradation. At the same time, the sibling status of O. daurica and O. curzoniae (Smith et al.,

136 GENERAL DISCUSSION 6.2. ECOSYSTEM IMPACT

1990) may imply that they show similar behaviours and O. curzoniae shows population dynamics of the group of burrowing pikas. But Yu et al. (1997) question the sibling status of the two species when using molecular rather than morphological traits to study the relationships among pika species. Thus, there remains considerable doubt if O. daurica can be considered a “burrowing” species in the sense of Smith (1988). If this species also exhibits aggressive and territorial behaviour in the research area, this will strengthen the arguments linking the predictability of the environment with territorial behaviour within the pikas and thus the applicability of the r/K-concept to pikas.

6.2 Ecosystem impact

Before deciding whether control actions should be taken against the Mongolian pika, its impact on the ecosystem must be investigated to find out the benefit or damage the pika causes. Putman (1989) proposes to define an animal pest as “any animal species or population which by its activities conflicts with Man’s interest to a level where the damage caused becomes of economic significance”, but at the same time reminds to include into the balance the cost of control and the repercussions of control on ecosystem services in the balance. Only after weighing this information, an economic injury level can be deduced (Begon et al., 1996). So the following questions must be answered: What damage does the pika cause? How can it be controlled? What are the repercussions of control?

Damage caused by pikas Reducing pika densities may result in more standing crop available for other herbivores. Even after including the positive eﬀect of burrows on ecosystem productivity, pika biomass removal always lowered the standing crop available (Table 5.9, p. 128).

Feasibility of control The results of the present study imply that the cost of controlling the Mongolian pika may be high. Territorial behaviour and density-dependent mortality (Section 4.2, p. 79) lead to relatively stable population densities controlled by the availability of burrows. Destroying burrows as a measure of controlling pikas as suggested for the rabbit in Aus- tralia by Eldridge and Simpson (2002) would be rather cost-intensive. This measure is impossible to use in a National Conservation Park and will not be accepted by the Mongolians, who for religious reasons do not like to break the soil surface. Additionally, the loosened soil will easily erode and pikas will ﬁnd it even easier to dig new burrows in the loosened soil. Increasing summer mortality by shooting or poisoning animals will not reduce densities, since mortality is high during the reproductive season anyway (Section 4.2, p. 79). Possible control measures may be increasing winter mortality or reducing fecundity in summer. Another measure proposed to reduce numbers of small mammal herbivores is to improve range condition. Shi (1983) shows that population density of Ochotona curzoniae was higher in areas with lower and less dense vegetation. Zhong et al. (1985) stress that small mammal activities are only detrimental for the range if it was initially

137 6.2. ECOSYSTEM IMPACT GENERAL DISCUSSION overgrazed by livestock and that improving the range is the best measure against damage caused by small mammals. Within the research area, grazing pressure is high and uniform (Stumpp et al., 2005), but signs of overgrazing have not been found (Miehe, 1996, 1998; Wesche et al., in press). This may be the reason, why the present study did not ﬁnd a strong eﬀect of livestock densities on pika burrow densities (Section 5.2, p. 124). Climatic conditions and not livestock control the amount of standing crop produced in the area (Retzer and Reudenbach, 2005; Wesche and Retzer, 2005). Ad- ditionally, higher standing crop in the year 2000 did not reduce pika densities. On the contrary, after the suite of years with favourable conditions for vegetation growth up to 2000, pika density was high relative to the following years (Chapter 3, p. 47). Thus, even if range condition would improve by limiting livestock grazing, pika densities are expected to stay the same in the research area. And, the other way round, even if pika numbers could be reduced, it would be questionable, if this would lead to higher amounts of standing crop. Section 5.1 on page 109 shows that individual pikas can remove varying amounts of biomass from the standing crop. Reducing pika numbers may thus result in more plant biomass stored away in burrows, and not in more standing crop available for livestock.

Repercussions To assess the repercussions on the ecosystem in the case of successfully controlling pikas, their function within the ecosystem must be considered. Function is here under- stood as “function as a service for humans” following Jax (2005). Three of the goods for humans provided by grassland are (a) food, forage and livestock, (b) biodiversity, and (c) tourism and recreation (White et al., 2000). Accordingly, the National Con- servation Parks in Mongolia distinguish three diﬀerent zones of protection, (a) a zone of “limited use” where livestock grazing is allowed, (b) a “special use” zone of nature protection, where neither tourism nor livestock are admitted, and (c) a “tourist” zone (Bedunah and Schmidt, 2000). a) Food, forage and livestock The impact of pika activity on standing crop is a function interesting from the perspective of pastoralists: one of the reasons why organic matter is more deeply distributed in steppe ecosystems is the abundance of soil fauna (Acton, 1992). Even in the case of the Brandt’s vole (Microtus brandtii), areas which have been “destroyed” by their burrowing activity are more productive in the long run (Weiner et al., 1982). In the present study, burrows could use available moisture better than steppe plots. Plant biomass was higher on burrow plots than on steppe plots under conditions of more precipitation (Figure 5.2, p. 115). In another study we show that the soil of burrows is richer in nutrients (Wesche et al., 2003, submitted). This is in accordance with other studies on pika burrows: Ochotona daurica loosens and improves the soil, biomass of roots and shoots is greater around burrow systems (Tsendzhav, 1985). Stumpp et al. (2005) conclude that pika activities conserve nutrients in the system, which are otherwise transported to water holes in form of livestock dung and subsequently used as fuel source by humans. After humans have burned the dung of livestock the nutrients are lost for the ecosystem. Lavrenko et al. (1993) distinguish between permanent and non-permanent burrows in Eurasian steppes, the later experiencing a succession towards steppe vegetation after abandonment. Although they classify pika as non-permanent, the present study

138 GENERAL DISCUSSION 6.2. ECOSYSTEM IMPACT

estimated the age of pika burrows to be higher than 120 years (Section 2.4.1, p. 26). It is thus questionable, if the impact of pikas on the soil is only local on the burrows themselves or also effects the steppe soil in the long run, e.g. by shifting burrow locations during centuries. At the same time, the plant profiting most from burrow conditions is Agropyron cristatum, which on burrows did not only show highest growth rates (Figure 5.3, p. 119), but also higher nutrient concentrations (Wesche et al., submitted), and additionally an earlier phenology and livestock prefer grazing on burrows (Retzer, 2004; Tsendzhav, 1985). Pikas prefer A. cristatum on burrows to that on steppe habitat (Table 5.6, p. 120), indicating that the plants on burrows may be easier to digest. The influence of feeding excavations by pikas acting as soil pockets to collect seeds, moisture, and organic matter has not been studied yet but may be important for dynamics of the vegetation (Section 5.1, p. 109, Eisenberg and Kinlaw, 1999). Although there is intense competition for food under drought conditions, livestock and pikas can coexist in the research area (Retzer and Reudenbach, 2005). The intensity of competition changes in time and space, so that actual competition for food probably is rare in the system. Even in the drought year, the herders did not complain about the pikas. In interviews, they did not show much interest in pikas (V. Retzer, pers. comm.).

b) Biodiversity and nature conservation Nature conservation aims at protect- ing the existing ecosystem (Bedunah and Schmidt, 2000), which self-evidently includes small mammals. The present study shows that human activities have little influence on the densities of either pikas or their burrows. Pika densities were mainly controlled by the availability of burrows (Chapter 4, p. 57) and burrow densities were mainly controlled by altitude, while livestock densities had only little effect on burrow densities (Section 5.2, p. 124). Pikas are thus seen as native part of the mountain steppe ecosystem, not as animals promoted by human activity. Apart from considerations about plant standing crop, pikas offer a range of other services to the ecosystem, including serving as prey and providing structures for other species. Many of the rare and endangered species in the GGS are birds of prey (Ta- ble 2.3, p. 18). Pikas serve as important prey base for these birds, since they are also available during winter (own observation and Smith et al., 1990). Their burrows serve as refuge and nesting site for other species. In the research area, Oenanthe isabellina has been observed to nest in pika burrows. Additionally, the higher altitudes of the park are designated as the core areas for nature protection, where human use is supposed to be inhibited (Miehe, 1996; Reading et al., 1999). Thus, competition with livestock should not be an issue in these areas. But excluding livestock from the higher altitudes may be impossible (Bedunah and Schmidt, 2000; Reading, 1996; Wesche et al., in press), since livestock is free roaming. Moreover the strengthening of pastoral economy is a declared aim of the management of the park. The park wants to protect moderately grazed steppe and desert ecosystems (Bedunah and Schmidt, 2000). Indeed, the feasibility of nature conservation depends on the good will of local communities (IIED, 1994) and pastoral use is the most sustainable form of human exploitation of steppe ecosystems (Sneath, 1998; White et al., 2000).

139 6.2. ECOSYSTEM IMPACT GENERAL DISCUSSION

c) Tourism The best way to ensure an interest of local people in the protection of their environment is to let them proﬁt directly from the protected goods, which can be achieved by tourism. Especially pikas could play a role here. Since they are cute and easy to observe, they can be nicely used to illustrate the interactions between the diﬀerent trophic layers of the ecosystem as well as social behaviour in animals.

The considerations put down in this chapter and the ﬁndings of this study on the life history strategy and the ecosystem impact of the species result in the conclusion that the Mongolian pika is no pest, but rather beneﬁts the pasture. So control actions are not advisable.

140 GENERAL DISCUSSION 6.3. FURTHER RESEARCH NEEDS

6.3 Further research needs

Of the several aspects touched in the present study, many deserve further investiga- tion. In the following, the most urgent questions resulting from the present study are accompanied with suggestions for feasible research projects:

Behaviour Aggressive behaviour and territoriality The relation of aggressive behaviour, territoriality, and mating system can be studied for Ochotona pallasi pricei as well as for Ochotona daurica. Individuals of both species can be marked and observed using a car as shelter for the observer.

Aggressive behaviour and productivity How is aggressive and territorial behaviour related to gradients of productivity? Ochotona pallasi behaviour shows different degrees of aggression. The subspecies in Kazakhstan is less aggressive than the that in the Mongolian Altai. The question is, whether differences in aggression can be explained by differences in productivity. Aggression can be sampled by observing focal individuals which do not have to be individually marked. Aggressive or affiliative behaviour towards any neighbours can be noted during a given period of time, e.g. 15 minutes. Focal individuals can be changed in random directions from the observer. Productivity of the vegetation can be sampled by measuring vegetation height and cover together with the density of pika burrows for different sites in the Altai mountains. Additionally, the amount of burrows presently occupied should be noted.

Ecosystem impact Burrow development The study area on which the burrows were mapped during this study should be mapped again in intervals of five to ten years. The next mapping would then be due in the year 2008. Following burrow development over a longer time can assess whether the influence of the burrowing activity of pikas is only local or if it effects a larger area, e.g. by shifting of burrow locations over time.

Feeding excavations What area is eﬀected by the feeding excavations? Are the resulting soil pockets successfully used by plant diaspores to establish themselves? What role do ants play in establishing microhabitats for germination? These questions can be addressed by permanently marking foraging pits and following them over time.

Commensals Given the approximate life-time of pika burrows, there are probably several species dependent on these reliable structures in the mountain steppes. Burrows should be investigated in various places in the Altai mountains to discern possible cases of coevolution between the burrowing pikas and other species living in their burrows.

141 6.3. FURTHER RESEARCH NEEDS GENERAL DISCUSSION

142 Bibliography

Acton, D. F., 1992. Grassland soils. In Coupland (1992b), pp. 25–54.

Allaby, M., ed., 1999. ”pest” A Dictionary of Zoology [online]. Oxford Reference Online, Oxford University Press, Oxford. Available from: Universit¨atsbibliothek Gießen. Last visited 15.2.2006, URL www.oxfordreference.com.

Barnett, A. and J. Dutton, 1995. Expedition ﬁeld techniques: Small mammals (excluding bats). Expedition Advisory Centre, London.

Barthel, H., 1990. Mongolei - Land zwischen Taiga und Wuste¨ . VEB H. Haack, Gotha, 220 pp.

Bedunah, D. J. and S. M. Schmidt, 2000. Rangelands of Gobi Gurvan Saikhan National Conservation Park, Mongolia. Rangelands, 22(4): 18–24.

Begon, M., J. L. Harper, and C. R. Townsend, 1996. Ecology: Individuals, Populations and Communities. Blackwell Science Ltd., Oxford, London, Edinbourgh.

Berryman, A. A., M. Lima, and B. A. Hawkins, 2002. Population regulation, emergent properties, and a requiem for density dependence. Oikos, 99: 600–606.

Bjørnstad, O. N., R. M. Nisbet, and J.-M. Fromentin, 2004. Trends and cohort resonant eﬀects in age-structured populations. Journal of Animal Ecology, 73: 1157–1167.

Bourli`ere, F., 1975. Introduction. In F. B. Golley, K. Petrusewicz, and L. Ryszkowski, eds., Small mammals: their productivity and population dynamics, vol. 5 of Interna- tional Biological Programme, pp. 1–23. Cambridge University Press, Cambridge.

Boyce, M. S., 1984. Restitution of r- and K-selection as a model of density-dependent natural selection. Annual Review of Ecology and Systematics, 15: 427–447.

Brockhaus, F. A., ed., 1999. ”pest” Brockhaus - Die Enzyklop¨adie. [online]. Bibli- ographisches Institut F. A. Brockhaus AG, Munzinger-Archiv GmbH, Mannheim, Ravensburg. Available online from: Universit¨atsbibliothek Gießen. Last visited 15.2.2006.

Burnham, K. P. and G. C. White, 2002. Evaluation of some random eﬀects methodology applicable to bird ringing data. Journal of Applied Statistics, 29(1-4): 245–264.

Butler, D., 1995. Zoogeomorphology: animals as geoporphic agents. Cambridge Uni- versity Press, Cambridge, MA.

143 BIBLIOGRAPHY

Carothers, A. D., 1973. Capture-recapture methods applied to a population with known parameters. Journal of Animal Ecology, 42: 125–146.

Caughley, G., 1977. Analysis of vertebrate populations. John Wiley and Sons, Toronto.

Chapman, J. A. and J. E. C. Flux, 1990a. Introduction and Overview of the Lago- morphs. In Chapman and Flux (1990b).

Chapman, J. A. and J. E. C. Flux, eds., 1990b. Rabbits, Hares and Pikas. Status Survey and Conservation Action Plan. IUCN, Gland, Switzerland.

Clauss, M. and J. Hummel, 2005. The digestive performance of mammalian herbivores: why big may not be that much better. Mammal Review, 35(2): 174–187.

Cormack, R. M., 1964. Estimates of survival from the sighting of marked animals. Biometrika, 51(3/4): 429–438.

Coupland, R. T., 1992a. Approach and generalizations. In Coupland (1992b), pp. 1–6.

Coupland, R. T., ed., 1992b. Natural grasslands. Introduction and Western Hemi- sphere. No. 8A in Ecosystems of the World. Elsevier, Amsterdam, 469 pp.

Dawson, M. R., 1967. Lagomorph History and the Stratigraphic Record. Essays in Paleontology and Stratigraphy, pp. 287–316.

Demin, E. P., 1962. [About the biology of the Mongolian pika]. Izvestia Irkutskogo. Nauchno-Issled. Protivochumnogo Ins. Sib. Dal’nego Vost., 24: 303–307. In Russian.

Ebright, J. R., T. Altantsetseg, and R. Oyungerel, 2003. Emerging Infectious Diseases in Mongolia. Emerging Infectious Diseases (serial online). Last visited 2005-12-14, URL http://www.cdc.gov/ncidod/EID/vol9no12/02-0520.htm.

Eisenberg, J. F. and A. Kinlaw, 1999. Introduction to the Special Issue: ecological signiﬁcance of open burrow systems. Journal of Arid Environments, 41: 123–125.

Eldridge, D. J. and R. Simpson, 2002. Rabbit (Oryctolagus cuniculus L.) impacts on vegetation and soils, and implications for management of wooden rangelands. Basic and Applied Ecology, 3: 19–29.

Ellis, J. E. and D. M. Swift, 1988. Stability of African pastoral ecosystems: Alternate paradigms and implications for development. Journal of Range Management, 41: 450–459.

Farnsworth, K. D., S. Focardi, and J. A. Beecham, 2002. Grassland-herbivore interactions: How do grazers coexist? American Naturalist, 159(1): 24–39.

Fernandez-Gimenez, M. E. and B. Allen-Diaz, 1999. Testing a non-equilibrium model of rangeland vegetation dynamics in Mongolia. Journal of Applied Ecology, 36: 871– 885.

Formozov, A. N., 1968. Adaptive modiﬁcations of behavior in mammals of the Eurasian steppes. Journal of Mammalogy, 47(2): 208–223.

144 BIBLIOGRAPHY

Formozov, N. A. and N. S. Proskurina, 1980. [Spatial structure of the settlements, and elements of the behaviour of closely related forms of pikas]. In [Rodents. Materials of the ﬁfth All-Union meeting], pp. 293–294. Nauka, Moscow-Leningrad. In Russian.

Grime, J. P., J. G. Hodgson, and R. Hunt, 1988. Comparative Plant Ecology. Allen & Unwin, London.

Grinnell, J., 1923. The burrowing rodents of California as agents in soil formation. Journal of Mammalogy, 4: 137–149.

Grunell, J. and J. R. Flowerdew, 1990. Live Trapping Small Mammals. A Practical Guide, vol. 3 of An Occasional Publication of the Mammal Society. 36 pp.

Hafner, D. J., 1994. Pikas and permafrost: Post-Wisconsin historical zoogeography of Ochotona in the southern Rocky Mountains, U.S.A. Arctic and Alpine Research, 26(4): 375–382.

Hairston, N. G., F. E. Smith, and L. B. Slobodkin, 1960. Community structure, population control, and competition. American Naturalist, 94: 421–25.

Hilbig, W., 1995. The vegetation of Mongolia. SPB Academic Publishing, Amsterdam.

Hilbig, W. and C. Opp, 2005. The eﬀects of anthropogenic impact on plant and soil cover in Mongolia. Exploration into the Biological Resources of Mongolia, 9: 163–178.

Hilborn, R., J. A. Redﬁeld, and C. Krebs, 1976. On the reliability of enumeration for mark and recapture census of voles. Canadian Journal of Zoology, 54(6): 1019–1024.

Holst, D. v., H. Hutzelmeyer, P. Kaetzke, M. Khaschei, and R. Sch¨onheiter, 1999. Social Rank, Stress, Fitness, and Life Expectancy in Wild Rabbits. Naturwissenschaften, 86: 388–393.

Holtmeier, F.-K., 1999. Tiere als ok¨ ologische Faktoren in der Landschaft. Arbeiten aus dem Institut fur¨ Landschaftsökologie Westfälische Wilhelms-Universität, 6: 1–348.

Huntly, N., 1991. Herbivores and the dynamics of communities and ecosystems. Annual Review of Ecology and Systematics, 22: 477.

Huntly, N. J., 1987. Inﬂuence of refuging consumers (pikas: Ochotona princeps) on subalpine meadow vegetation. Ecology, 68(2): 274–283.

IIED, 1994. Whose Eden? An overview of community approaches to wildlife management. International Institute for Environmental Development, London.

Illius, A. W. and T. G. O. Connor, 1999. On the relevance of nonequilibrium concepts to arid and semiarid grazing systems. Ecological Applications, 3: 798–813.

Jax, K., 2005. Function and ”functioning” in ecology: what does it mean? Oikos, 111(3): 641–648.

Jolly, G. M., 1965. Explicit estimates from capture-recapture data with both death and immigration-stochastic model. Biometrika, 52(1-2): 225–247.

145 BIBLIOGRAPHY

Jones, C. G., J. H. Lawton, and M. Shachak, 1994. Organisms as ecosystem engineers. Oikos, 69: 373–386.

Jones, C. G., J. H. Lawton, and M. Shachak, 1997. Positive and negative eﬀects of organisms as physical ecosystem engineers. Ecology, 78: 1946–1957.

Jouquet, P., J. Dauber, J. Lagerl¨of, P. Lavalle, and M. Lepage, 2005. Soil invertebrates as ecosystem engineers: Intended and accidental eﬀects on soil and feedback loops. Applied Soil Ecology.

Kaitala, V. and E. Ranta, 2001. Is the impact of environmental noise visible in the dynamics of age-structured populations? Proceedings of the Royal Society London, Series B, 268: 1769–1774.

Kawamichi, T., 1968. Winter Behaviour of the Himalayan Pika, Ochotona roylei. Jour. Fac. Sci. Hokkaido Univ. Ser. VI, Zool., 16: 582–594.

Kawamichi, T., 1971. Annual Cycle of Behaviour and Social Pattern of the Japanese Pika, Ochotona hyperborea yesoensis. Jour. Fac. Sci. Hokkaido Univ. Ser. VI, Zool., 18(1): 173–185.

Kawamichi, T. and J. Liu, 1990. Capturing and marking pikas (Ochotona) with systematic ear clipping patterns. Journal of the Mammal Society Japan, 15(1): 39–43.

Kinlaw, A., 1999. A review of burrowing by semi-fossorial vertebrates in arid environments. Journal of Arid Environments, 41: 127–145.

Kniazev, A. V. and A. B. Savinetski, 1988. Changes in the populations of small mammals of the Tsaagan-Bogdo ridge (Transaltai Gobi) in the late Holocene. Zo- ologitsheskii jurnal, 67: 297–300. In Russian.

Kniazev, A. V., A. B. Savinetski, and V. T. Sokolovskaya, 1986. Population dynamics in phytophagous mammals in the Transaltai Gobi in the middle and late Holocene. Zoologitsheskii jurnal, 65: 247–258. In Russian.

Komonen, M., A. Komonen, and A. Otgonsuren, 2003. Daurian pikas (Ochotona daurica) and grassland condition in eastern Mongolia. Journal of Zoology, 259: 281–288.

Koper, N. and R. J. Brooks, 1998. Population-size estimators and unequal catchability in painted turtles. Canadian Journal of Zoology, 76: 458–465.

Krebs, C. J., 1985. Ecology: The Experimental Analysis of Distribution and Abundance. Harper and Row, New York.

Krebs, C. J., 1999. Ecological methodology. Benjamin/Cummings, Menlo Park, Calif.

Krylova, 1974. [Reproduction of the Mongolian pika in the southwestern part of the Tuva ASSR]. Biol. Nauk, 17: 31–37. In Russion.

Krylova, T. W., 1973. [Age compositon of the Mongolian pika, Ochotona pricei population]. Zoologitsheskii jurnal, 53: 1422–1425. In Russian.

146 BIBLIOGRAPHY

Lavrenko, E. M., Z. V. Karamysheva, I. V. Borisova, T. A. Popova, N. P. Guricheva, and R. I. Nikulina, 1993. Steppes of the former Soviet Union and Mongolia. In R. T. Coupland, ed., Natural Grasslands. Eastern Hemisphere and R´esum´e, pp. 3–57. Elsevier, Amsterdam.

Lazarev, B. V., 1968. [The ecology of Pallas’ pika in the Altai mountains]. Izvestia Irkutskogo. Nauchno-Issled. Protivochumnogo Ins. Sib. Dal’nego Vost., 27: 17–22. In Russian.

Lebreton, J. D., K. P. Burnham, J. Clobert, and D. R. Anderson, 1992. Modelling survival and testing biological hypotheses using marked animals: a uniﬁed approach with case studies. Ecological Monographs, 62(1): 67–118.

Leirs, H., 1994. Population ecology of Mastomys natalensis (Smith, 1834). Implications for rodent control in Africa, vol. 35 of Agricultural Edition. Belgian Administration for Development Cooperation, Brussels, 268 pp.

Leont’ev, A. N., 1968. [Reproduction of the Daurian Pika in relation to ﬂuctuations in its population size]. Izv. Irkutsk. Nauchno-Issled. Protivochumnogo Ins. Sib. Dal’nego Vost., 27: 23–28. In Russian.

Lutton, L. M., 1975. Notes on territorial behavior and response to predators of the pika, Ochotona princeps. Journal of Mammalogy, 56(1): 231–234.

MacArthur, R. A. and L. C. H. Wang, 1973. Physiology of thermoregulation in the pika, Ochotona princeps. Canadian Journal of Zoology, 51: 11–16.

MacArthur, R. H. and E. O. Wilson, 1967. The Theory of Island Biogeography. Prince- ton University Press, Princeton, NJ.

Madhusudan, M. D., 2004. Recovery of wild large herbivores following livestock decline in a tropical Indian wildlife resource. Journal of Applied Ecoloy, 41: 858–869.

Manning, T., W. D. Edge, and J. O. Wolﬀ, 1995. Evaluating population-size estimators: an empirical approach. Journal of Mammalogy, 76: 1149–1158.

Mares, M. A., M. R. Willig, and N. A. Bitar, 1980. Home range size in eastern chipmunks, Tamias striatus, as a function of number of captures: statistical biases of inadequate sampling. Journal of Mammalogy, 61: 661–669.

Matsumoto, T., M. Kosaka, M. Saito, A. Sakai, T. Matsuzaki, S. Ganzorig, N. Ohwatari, M. Shimazu, and T. Nomura, 1995. A ﬁeld study of telemetry-recording of the body temperature in wild Mongolian pikas. Tropical Medicine, 37(3): 93–98.

Mcintire, E. J. B. and D. S. Hik, 2002. Grazing history versus current grazing: leaf demography and compensatory growth of three alpine plants in response to a native herbivore (Ochotona callaris). Journal of Ecology, 90: 348–359.

Miehe, S., 1996. Vegetationskundlich-ökologische Untersuchungen, Auswahl und Ein- richtung von Dauerprobeflächen fur¨ Vegetationsmonitoring im Nationalpark Gobi- Gurvan-Saikhan. Tech. rep., GTZ Nature Conservation and Buffer Zone Develop- ment Project. Unpublished report.

147 BIBLIOGRAPHY

Miehe, S., 1998. Ans¨atze zu einer Gliederung der Vegetation im Nationalpark Gobi- Gurvan-Saikhan. Tech. rep., GTZ Nature Conservation and Buﬀer Zone Develop- ment Project. Unpublished report.

Millar, J. S., 1972. Characteristics and ecological signiﬁcance of hay piles of pikas. Mammalia, 36(4): 657–668.

Millar, J. S., 1973. Evolution of litter-size in the pika, Ochotona princeps (Richardson). Evolution, 27: 134–143.

Millar, Z. and F. C. Zwickel, 1972. Determination of age, agestructure and mortality of the pika, Ochotona princeps (Richardson). Canadian Journal of Zoology, 50: 229–232.

Min´aˇr, J. and K. H˚urka, 1980. Parasit¨are Dipteren (Insecta, Diptera: Hypodermati- dae, Hippoboscidae, Nycteribiidae) aus der Mongolei. Mitteilungen des zoologischen Museums Berlin, 56(2): 187–189. In German.

Monkhzul, T., 2005. Breeding behavior of the Mongolian Pika (Ochotona pallasi in the Gobi Gurvan Saykhan mountains, Mongolia. Exploration into the Biological Resources of Mongolia, 9: 45–51.

Morrison, P. R. and W. J. Teitz, 1956. Observation on food consumption and preference in four Alaskan mammals. Arctic, 9: 52–57.

Muller,¨ F.-V., 1999. Die Wiederkehr des mongolischen Nomadismus. R¨aumliche Mo- bilit¨at und Existenssicherung in einem Transformationsland. Abhandlungen - An- thropogeographie Institut fur¨ Geographische Wissenschaften FU Berlin, 60: 11–46.

Naumov, R. L., 1974. Ecology of Ochotona alpina in the west Sayan. Zoologitsheskii jurnal, 53: 1524–1529. In Russian.

Nikol’skii, A. A. and T. D. Mukhamediev, 1997. [Altai pika (Ochotona alpina Pallas, 1773) square of individual ranges]. Byulleten’ Moskovskogo Obshchestva Ispytatelei Prirody Otdel Biologicheskii, 102(1): 20–28. In Russian, English summary.

Ognev, S. I., 1940. Mammals of the USSR and adjacent countries. Vol. 4: Rodents. Izv. Akademi Nauk USSR, Moscow.

Oksanen, L., S. Fretwell, J. Arruda, and P. Niemela, 1981. Exploitation ecosysems in gradients of promary productivity. American Naturalist, (118): 240–261.

Oksanen, L. and T. Oksanen, 1990. Long-term microtine dynamics in North Fennoscan- dian tundra: is there a case for a mustelid-microtine limit cycle? In Predator-prey dynamics in small mammals along gradients of primary productivity, p. 32. Ume˚a, Sweden. Phd thesis.

Okunev, L. P. and C. B. Zonov, 1980. [Ecological adaptation of Ochotona pricei to mountainous-steppe landscapes]. Ecologia, 6: 61–66. In Russian.

Olﬀ, H. and M. E. Ritchie, 1998. Eﬀects of herbivores on grassland plant diversity. Trends in Ecology and Evolution, 13: 261–265.

148 BIBLIOGRAPHY

Owen-Smith, N., 2005. Incorporating fundamental laws of biology and physics into population ecology: the metaphysiological approach. Oikos, 111(3): 611–615.

Parmenter, R. R., T. L. Yates, D. R. Anderson, K. P. Burnham, J. L. Dunnum, A. B. Franklin, M. T. Friggens, B. C. Lubow, M. Miller, G. S. Olson, C. A. Parmenter, J. Pollard, E. Rexstad, T. M. Shenk, T. R. Stanley, and G. C. White, 2003. Small- mammal density estimation: a ﬁeld comparison of grid-based vs. web-based density estimators. Ecological Monographs, 73(1): 1–26.

Peacock, M. M., 1997. Determining natal dispersal patterns in a population of North American pikas (Ochotona princeps) using direct mark-resight and indirect genetic methods. Behavioral Ecology, 8(3): 340–350.

Peacock, M. M. and C. Ray, 2001. Dispersal in pikas (Ochotona princeps): combining genetic and demographic approaches to reveal spatial and temporal patterns. In J. Clobert, E. Danchin, A. A. Dhondt, and J. D. Nichols, eds., Dispersal, pp. 43–56. Oxford University Press, Oxford, New York.

Petersen, C. G. J., 1896. The yearly immigration of young plaice into the Limfjord from the German Sea. Danish Biological Station Report, 6: 1–48.

Pianka, E. R., 1970. On r- and K-selection. American Naturalist, 104: 592–597.

Power, M. E. and et al, 1996. Challenges in the quest for keystones. BioScience, 46: 609–620.

Proskurina, N. S., N. A. Formozov, and D. G. Derviz, 1985. A comparative analysis of spatial structure of colonies in three forms of pikas: Ochtona daurica, O. pallasi pallasi, O. pallasi pricei(Lagomorpha, Lagomyidae). Zoologitsheskii jurnal, 64(11): 1695–1701. In Russian, English summary.

Puget, 1971. Ochotona r. rufenscens (Gray, 1842) en Afghanistan et son elevage en captivite. Mammalia, pp. 26–37.

Puget, A. and C. Gouarderes, 1974. Weight gain of the Afghan Pika (Ochotona rufenscens rufescens) from birth to 19 weeks of age, and during gestation. Growth, 38: 117–129.

Putman, R. J., 1989. Mammals as pests: Introduction. In R. J. Putman, ed., Mammals as pests, p. 1. Chapman and Hall, London.

Putman, R. J., 1995. Ethical considerations and animal welfare in ecological ﬁeld studies. Biodiversity and Conservation, 4: 903–915.

Quinn, G. P. and M. J. Keough, 2002. Experimental Design and Data Analysis for Biologists. University Press, Cambridge.

R Development Core Team, 2004. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. Last visited 30.3.2005, URL http://www.R-project.org.

149 BIBLIOGRAPHY

Reading, R. P., 1996. Biological Assessment of Three Beauties of the Gobi National Conservation Park II: Spring 1996 . German Technical Advisory Group’s (GTZ) Nature Conservation and Buﬀer Zone Development Project. Unpublished report.

Reading, R. P., S. Amgalanbaatar, and L. Lhagvasuren, 1999. Biological assessment of Three Beauties of the Gobi Naional Park, Mongolia. Biodiversity and Conservation, 8: 1115–1137.

Reading, R. P., H. M. Mix, B. Lhagvasuren, C. Feh, D. Kane, and S. Dalamtseren, 2001. Status and distribution of khulan (Equus hemonius) in Mongolia. Journal of Zoology, 254: 381–389.

Retzer, V., 2004. Carrying capacity and forage competition between livestock and a small mammal, the Mongolian Pika (Ochotona pallasi) in a non-equilibrium ecosystem. G¨orich & Weierh¨auser Verlag, Marburg.

Retzer, V., 2005. Facts from a year of drought: forage competition between livestock and the Mongolian pika (Ochotona pallasi) and its eﬀects on livestock densities and body condition. Exploration into the Biological Resources of Mongolia, 9: 157–162.

Retzer, V., submitted. Forage competition between livestock and Mongolian Pika (Ochotona pallasi) in southern Mongolian mountain steppes. Basic and Applied Ecology.

Retzer, V. and K. Nadrowski, 2002. Livestock and small mammal grazing in the mountain steppe of the Gobi Gurvan Saikhan National Park of Mongolia. In T. Chuluun and D. S. Ojima, eds., Fundamental Issues Aﬀecting Sustainability of the Mongolian Steppe, pp. 192–197. IISCN, Ulaanbaatar.

Retzer, V., K. Nadrowski, and G. Miehe, in press. Variation of precipitation and its eﬀect on phytomass production and consumption by livestock and large wild herbivores along an altitudinal gradient during a drought, South-Gobi, Mongolia. Journal of Arid Environments.

Retzer, V. and C. Reudenbach, 2005. Modelling the carrying capacity and coexistence of pika and livestock in the mountain steppe of the South Gobi, Mongolia. Ecological Modelling, 189: 89–104.

Rothshild, E. V. and A. G. Derevshchikov, 1994. [Ecologic-geochemical conditions for the plague natural focality manifestation in the Altai mountains]. Byulleten’ Moskovskogo Obshchestva Ispytatelei Prirody Otdel Biologicheskii, 99(6): 15–22.

Samjaa, R., U. Z¨ophel, and J. Peterson, 2000. The impact of the vole Microtus brandtii on Mongolian steppe ecosystems. Marburger Geographische Schriften, 135: 346–360.

Sammul, M., L. Oksanen, and M. Mägi, 2006. Regional effects on competition- productivity relationship: a set of field experiments in two distant regions. Oikos, 112: 138–148.

Schaller, G., 2000. Wildlife of the Tibetan Steppe. University of Chicago Press, Chicago, 373 pp.

150 BIBLIOGRAPHY

Schauer, J., 1987. Remarks on the construction of burrows of Ellobius talpinus, Myospalax aspalax and Ochotona daurica in Mongolia and their eﬀect on the soil. Folia Zoologica, 36(4): 319–326.

Schnabel, Z. E., 1938. The estimation of the total ﬁsh population of a lake. American Mathematics Monthly, 45: 348–352.

Schumacher, F. X. and R. W. Eschmeyer, 1943. The estimate of ﬁsh population in lakes or ponds. Journal of the Tennessee Academy of Science, 18: 228–249.

Schultz, 1995. Die Okozonen¨ der Erde. Ulmer, Stuttgart, 535 pp.

Scott Mills, L., M. E. Soul´e, and M. F. Doak, 1993. The key-stone species concept in ecology and conservation. Animal Conservation, 2: 235–240.

Seber, G. A. and C. J. Schwarz, 1999. Estimating animal abundance: Review III. Statistical Science, 14: 427–456.

Seber, G. A. F., 1965. A note on the multiple-recapture census. Biometrika, 52(1/2): 249–259.

Shi, Y., 1983. On the inﬂuence of rangeland vegetation to the density of Plateau pikas (Ochotona curzoniae). Acta Theriologica Sinica, 3: 181–187.

Shubin, I. G. [The dynamics of the age structure of the tree-creeper (Ochotona pusilla and Ochotona pallasi)]. Byulleten’ Moskovskogo Obshchestva Ispytatelei Prirody Ot- del Biologicheskii, 71: 27.

Skalski, J. R., A. Hoffmann, and S. G. Smith, 1993. Testing the significance of individual- and cohort-level covariates in animal survival studies. In J.-D. Lebre- ton and P. M. North, eds., Marked Individuals in the Study of Bird Populations, pp. 9–28. Birkhäuser-Verlag, Basel, Switzerland.

Smirnov, P. K., 1974. Biotopic distribution and territorial relationship of the steppe and Pallas’s pikas in the sympatric zone of their ranges. Byulleten’ Moskovskogo Obshchestva Ispytatelei Prirody Otdel Biologicheskii, 75: 72–80. In Russian, English summary.

Smith, A. T., 1987. Population Structure of Pikas: Dispersal versus Philopatry. In D. Chepko-Sade and Z. T. Halpin, eds., Mammalian Dispersal Patterns, pp. 128–142. The University of Chicago Press, Chicago, London.

Smith, A. T., 1988. Patterns of pika (Genus Ochotona) life history variation. In M. Boyce, ed., Evolution of life histories of mammals, pp. 233–256. Yale University Press, New Haven, London.

Smith, A. T. and J. M. Foggin, 1999. The Plateau Pika (Ochotona curzoniae) is a keystone species for biodiversity on the Tibetan plateau. Animal Conservation, 2: 235–240.

Smith, A. T., A. N. Formozov, , R. S. Hoﬀmann, C. Zheng, and M. A. Erbajeva, 1990. The pikas. In Chapman and Flux (1990b), pp. 14–60.

151 BIBLIOGRAPHY

Smith, A. T. and W. X. Gao, 1991. Social Relations of adult Black-Lipped Pikas (Ochotona curzoniae). Journal of Mammalogy, 72(2): 231–247.

Smith, A. T., H. J. Smith, X. Wang, X. Yin, and J. Liang, 1986. Social behavior of the steppe-dwelling blacklipped pika (Ochotona curzoniae). Acta Theriologica Sinica, 6(1).

Sneath, D., 1998. State Policy and Pasture Degradation in Inner Asia. Science, 281(21): 1147.

Sokolov, V. E. and V. N. Orlov, 1980. Otpredelitel mlekupitajushikh Mongolskii Narodnii Respubliki. Nauka, Moscow, 352 pp. In Russian.

Song, L. J. and L. Q. Fen, 1996. Variation of thermogenesis in Plateau pika (Ochotona curzoniae) during cold acclimation and de-cold acclimation. Acta Zoologica Sinica, 42(4): 377–385.

Stearns, S. C., 1977. The evolution of life history traits. Annual Review of Ecology and Systematics, 8: 145–171.

Stearns, S. C., 1992. The Evolution of Life Histories. Oxford University Press, Oxford.

Stenseth, 1989. How to control pest species. Journal of Applied Ecology, 18: 773–791.

Stenseth, N. C., A. Mysterud, G. Ottersen, J. W. Hurrell, K.-S. Chan, and M. Lima, 2002. Ecological eﬀects of climatic ﬂuctuations. Science, 297: 1292–1296.

Stubbe, 1971. Zur S¨augetierfauna der Mongolei III. Taiga - Pfeifhasen, Ochotona alpina hyperborea (Pallas, 1811), aus dem Chentej. Mitteilungen des Zoologischen Museums Berlin, 47(2): 349–356.

Stumpp, M., K. Wesche, V. Retzer, and G. Miehe, 2005. Impact of grazing livestock and distance from water points on soil fertility in southern Mongolia. Mountain Research and Development, 25(3): 245–252.

Sullivan, S. and R. Rohde, 2002. On non-equilibrium in arid and semi-arid grazing systems. Journal of Biogeography, 29: 1595–1618.

Thenius, E., 1980. Grundzuge¨ der Faunen- und Verbreitungsgeschichte der S¨augetiere. Gustav Fischer Verlag, Stuttgart, New York.

Titlyanova, A. A., A. A. Romanova, N. P. Kosykh, and T. Mironycheva, 1999. Pattern and process in above-ground and below-ground components of grassland ecosystems. Journal of Vegetation Science, 10: 307–320.

Tsendzhav, D., 1976. [About the bioecology of the Daurian pika in the Khangay]. Eronkhii ba Corilyn Biologiin Khureelengiin Erdem Shindshilgeenii Buteel, 11: 106– 114. In Mongolian, Russian summary.

Tsendzhav, D., 1977. [Population structure of the Daurian pika (Ochotona daurica)]. Eronkhii ba Corilyn Biologiin Khureelengiin Erdem Shindshilgeenii Buteel, 12: 73– 79. In Mongolian, Russian summary.

152 BIBLIOGRAPHY

Tsendzhav, D., 1985. [The role of the Daurian pika (Ochotona daurica Pallas, 1776) in the biogeocenosis of Eastern Khangai]. Ulaanbaatar, 31 pp. Phd thesis. In Mongolian, Russian summary.

UB Post. Ulaanbaatar. An English speaking newspaper.

Vorontsov, N. N. and E. U. Ivanitskaya, 1973. Comparative karyology of north Palaearc- tic pikas (Ochotona, Ochotonidae, Lagomorpha). Caryologia, 26: 213–223.

Wang, D. and Z. Wang, 1996. Seasonal variations in thermogenesis and energy require- ments of Plateau pikas Ochotona curzoniae and root voles Microtus oeconomus. Acta Theriologica Sinica, 41(3): 225–236.

Wang, M., W. Zhong, and X. Wan, 2000. A study of home range of Daurian pika (Ochotona daurica) through telemetry. Acta Theriologica Sinica, 20(2): 116–122.

Wang, X., 1990. A study on the mating season and the pattern on copulation behavior in Plateau pika. Acta Theriologica Sinica, 10(1): 60–65.

Wang, X. and K. Dai, 1990. A study on the breeding area and the territorial behavior in Plateau pika (Ochotona curzoniae). Acta Theriologica Sinica, 10(3): 203–209.

Wang, X. and A. T. Smith, 1988. On the natural winter mortality of the Plateau pika (Ochotona curzoniae). Acta Theriologica Sinica, 8(2): 152–156.

Wehner, R. and W. Gehring, 1990. Zoologie. Georg Thieme Verlag, Stuttgart.

Weiner, J. and A. Gorecki, 1982. Small mammals and their habitat in the arid steppe of central Mongolia. Polish Ecological Studies, 8: 7–21.

Weiner, J., A. Gorecki, and K. Perzanowski, 1982. Standing crop and standing above ground production of vegetation in arid Mongolian steppe with Caragana. Polish Ecological Studies, 8: 7–21.

Wesche, K., S. Miehe, and G. Miehe, in press. Plant communities of the Gobi Gurvan Saykhan National Park (South Gobi Aimag, Mongolia). Candollea.

Wesche, K., K. Nadrowski, D. Enkhjargal, and R. Undrakh, 2003. Small mammal burrows as special habitats in Southern Mongolian mountain steppes. Verhandlungen der Gesellschaft fur¨ Okolo¨ gie, 33: 356.

Wesche, K., K. Nadrowski, and V. Retzer, submitted. Habitat engineering under dry conditions: The impact of pikas (Ochotona pallasi) on southern Mongolian steppes. Journal of Vegetation Science.

Wesche, K. and V. Retzer, 2005. Is degradation a major problem in semi-desert environments of Gobi region in southern Mongolia? Exploration into the Biological Resources of Mongolia, 9: 133–146.

Wesche, K. and K. Ronnenberg, 2004. Phytosociological aﬃnities and habitat preferences of Juniperus sabina (L.) and Artemisia santolinifolia (Turz. ex Bess.) in mountain sites of the south-eastern Gobi Altay, Mongolia. Feddes Repertorium, 115(7-8): 585–600.

153 BIBLIOGRAPHY

Wesche, K., K. Ronnenberg, and I. Hensen, 2005. Lack of sexual reproduction within mountain steppe populations of the clonal shrub Juniperus sabina L. in semi-arid southern Mongolia. Journal of Arid Environments, 63: 390–405.

Wesche, K. and H. von Wehrden, 2002. Vegetation types of the Gobi Gurvan Saikhan National Park (South Gobi Aimag, Mongolia) – Classiﬁcation and spatial distribution. GTZ Nature Conservation and Buﬀer Zone Development Project, Ulaanbaatar, 62 pp. Unpublished report.

White, G. C., D. R. Anderson, K. P. Burnham, and D. L. Otis, 1982. Capture-recapture and removal methods for sampling closed populations. Los Alamos National Labo- ratory, Los Alamos, New Mexico, USA.

White, G. C. and K. P. Burnham, 1999. Program MARK: Survival estimation from populations of marked animals. Bird Study, 46(Supplement): 120–138.

White, R. P., S. Murray, and M. Rohweder, 2000. Pilot Analysis of Global Ecosystems. Grassland Ecosystems. World Resource Institute, Washington, DC.

White, T. C. R., 2004. Limitation of populations by weather-driven changes in food: a challenge to density-dependent regulation. Oikos, 105(3): 664–666.

Whitford, W. G. and F. R. Kay, 1999. Biopedturbation by mammals: a review. Journal of Arid Environments, 41: 203–230.

Wiens, N., 1984. On understanding a non-equilibrium world: myth and reality in community patterns and processes. In D. R. Strong, D. Simberloﬀ, L. Abele, and A. B. Thistle, eds., Ecological communities: Conceptual issues and processes, pp. 339–457. Princeton University Press.

Wilson, K. R. and D. R. Anderson, 1985. Evaluation of two density estimators of small mammal population size. Journal of Mammalogy, 66: 13–21.

Ye, R. and J. Liang, 1989. Study on the growth and development of Plateau pika under the condition of artiﬁcial feeding. Acta Theriologica Sinica, 9(2): 110–118.

Yu, N., C. Zheng, and L. Shi, 1997. Variation in mitochondrial DNA and phylogeny of six species of pikas (Ochotona). Journal of Mammalogy, 78(2): 387–396.

Zevegmid, D., 1975. Biologie der Pfeifhasen (Ochotonidae) in der Mongolischen Volk- srepublik. Mitteilungen des Zoologischen Museums Berlin, 51: 41–53.

Zhong, W., Q. Zhou, and C. Sun, 1985. The basic characteristics of the rodent pests on the pasture in Inner Mongolia and the ecological strategies of controlling. Acta Theriologica Sinica, 5: 241–249.

Zonov, G. B., L. P. Okunev, and I. K. Mashkovsky, 1983. The use of air holes in snow by small mammals. Zoologidsheskii jurnal, 62(12): 1863–1867.

154 CURRICULUM VITAE

Karin Nadrowski

geboren 31.12.1971 in Neumunster,¨ Schleswig-Holstein

1982 – 1992 Gymnasium Klaus-Groth Schule in Neumunster.¨

1988/89 Stipendiatin des Parlamentarischen Patenschaftsprogramms: aus Austauschschulerin¨ in den USA, High-School Diploma.

1992 Abitur, Leistungsf¨acher: Mathematik und Physik, weitere Prufungsf¨¨ acher: Englisch, Geschichte;

1992/1993 Freiwilliges Ok¨ ologisches Jahr auf dem Darß, Jugendherberge Born.

1993-1995 Grundstudium der Biologie an der Philipps-Universit¨at in Marburg.

1995-1999 Hauptstudium Biologie an der Friedrich Schiller Universit¨at in Jena, Thuringen.¨ Hauptfach Ok¨ ologie, Nebenf¨acher Mathematische Biolo- gie, Zoologie, Soziologie.

1999 Diplomarbeit zur “A Source Sink System with Directed Resource Flow - Pattern of Vegetation in a Semi-arid Shrubland.” Dazu ein 5-monatiger Aufenthalt in der Wuste¨ Negev am Mitrani Center for Desert Ecology, Israel.

1999-2000 Forschungsauftrag am Max-Planck-Institut fur¨ Biogeochemie: Einbau einer Beweidungsroutine in ein globales Vegetationsmodell (LPJ).

2000-2006 Doktorarbeit am Fachbereich Geographie der Philipps Universit¨at Marburg. Thema: “Life history strategy and ecosystem impact of a dominant small mammal herbivore in a mountain steppe”.

2000-2001 Aufenthalt in der Mongolei. Aufbau der Forschungsstation und Fel- darbeit im Gobi-Altai.

2002 Erneuter Aufenthalt in der Mongolei.

2005 Geburt meiner Tochter Janis.

155